Navigation Menu

Skip to content

Instantly share code, notes, and snippets.

@puyiran

puyiran/logicbelief-master-puwork_.gitignore

Created July 12, 2020 14:25
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save puyiran/de31939aac36c8daeed215c278d10d5e to your computer and use it in GitHub Desktop.
Save puyiran/de31939aac36c8daeed215c278d10d5e to your computer and use it in GitHub Desktop.
Download ZIP
This gist exceeds the recommended number of files (~10). To access all files, please clone this gist.
Raw
logicbelief-master-puwork_.gitignore
*/.ipynb_checkpoints/*
.ipynb_checkpoints/*
db
db8
sendsms.py
__pycache__/*
nohup.out
venv
*.swp
buy
sell
Raw
logicbelief-master-puwork_.idea_.gitignore
# Default ignored files
/workspace.xml
Raw
logicbelief-master-puwork_.idea_inspectionProfiles_profiles_settings.xml
<component name="InspectionProjectProfileManager">
<settings>
<option name="USE_PROJECT_PROFILE" value="false" />
<version value="1.0" />
</settings>
</component>
Raw
logicbelief-master-puwork_.idea_logicbelief-master-puwork.iml
<?xml version="1.0" encoding="UTF-8"?>
<module type="PYTHON_MODULE" version="4">
<component name="NewModuleRootManager">
<content url="file://$MODULE_DIR$" />
<orderEntry type="inheritedJdk" />
<orderEntry type="sourceFolder" forTests="false" />
</component>
</module>
Raw
logicbelief-master-puwork_.idea_modules.xml
<?xml version="1.0" encoding="UTF-8"?>
<project version="4">
<component name="ProjectModuleManager">
<modules>
<module fileurl="file://$PROJECT_DIR$/.idea/logicbelief-master-puwork.iml" filepath="$PROJECT_DIR$/.idea/logicbelief-master-puwork.iml" />
</modules>
</component>
</project>
Raw
logicbelief-master-puwork_.idea_vcs.xml
<?xml version="1.0" encoding="UTF-8"?>
<project version="4">
<component name="VcsDirectoryMappings">
<mapping directory="$PROJECT_DIR$" vcs="Git" />
</component>
</project>
Raw
logicbelief-master-puwork_balance_top.py
from common import *
candidates = set()
data = pro.query('stock_basic', exchange='', list_status='L', fields='ts_code, industry')
with open("watch_industries.txt") as f:
content = f.readlines()
industries = set()
for x in content:
industries.add(x.strip())
for index, basic in data.iterrows():
for basic['industry'] in industries:
candidates.add(basic['ts_code'])
#print(dichan)
period = '20200331'
income = pro.income_vip(period = period)
net_profit_dict = {}
for index, cash in income.iterrows():
code = cash['ts_code']
if not (code in candidates):
continue
ebit = float(cash['ebit'])
if math.isnan(ebit) or ebit is None or ebit < 100000000:
continue
# if code == '600012.SH':
# print("%s %.2f" %(code, ebit))
net_profit_dict[code] = ebit
data = pro.balancesheet_vip(period = period)
ratio_dict = {}
for index, balance in data.iterrows():
code = balance['ts_code']
if code in net_profit_dict:
accounts_receiv = float(balance['accounts_receiv'])
if math.isnan(accounts_receiv):
accounts_receiv = 0
acct_payable = float(balance['acct_payable'])
if math.isnan(acct_payable):
continue
ratio_dict[code] = accounts_receiv /acct_payable
sorted_ratios = sorted(ratio_dict.items(), key=lambda d: d[1])
n = 0
column_str = """date, code, value, name"""
queries = []
for (k, v) in sorted_ratios:
n += 1
if n > 10:
break
print(k)
print(v)
Raw
logicbelief-master-puwork_buy.py
import tushare as ts
import warnings
import threading
import time
import sys
print(ts.__version__)
warnings.simplefilter(action='ignore', category=FutureWarning)
import easyquotation
quotation = easyquotation.use("timekline")
from send_wx import *
from common import *
from collections import defaultdict
con = get_conn("db8")
max_date = get_max_date()
print(max_date)
def obtain_list_of_db_tickers(table):
cur = con.cursor()
query_str = "select distinct code from %s where date = '%s'" % (table, max_date)
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
class BuyThread(threading.Thread):
def __init__(self, index, thread_count, model, close_dicts, volume_dicts, is_signal):
threading.Thread.__init__(self)
self.index = index
self.thread_count = thread_count
self.position_table = "%s_position" %(model)
self.model = model
self.close_dicts = close_dicts
self.volume_dicts = volume_dicts
self.is_signal = is_signal
def run(self):
print("Starting thread %d \n" % self.index)
buy(self.index, self.thread_count, self.model, self.position_table, self.close_dicts, self.volume_dicts, self.is_signal)
def buy(idx, thread_count, model, position_table, close_dicts, volume_dicts, is_signal):
count_dicts = defaultdict(int)
buy_dicts = defaultdict(int)
buy_queries = []
bought = set()
while not is_after_hour():
if not is_in_trade_time():
time.sleep(60)
continue
i = -1
for stock in stocks:
i += 1
if (i % thread_count) != idx:
continue
# print(stock)
if stock in bought:
continue
code = stock[:-3]
sc = stock[-2:].lower()
try:
key = "%s%s.js" % (sc, code)
obj = quotation.real([code])[key]['time_data']
total_len = len(obj)
for cur_len in range(count_dicts[stock] - 1, total_len):
if cur_len < 3:
continue
cur_time = obj[cur_len][0]
if buy_dicts[stock] == int(cur_time):
continue
cur_price = float(obj[cur_len][1])
cur_volume = float(obj[cur_len][2])
last_volume = float(obj[cur_len - 1][2])
sec_volume = cur_volume - last_volume
avg_vol = cur_volume / (1 + cur_len)
last_avg_vol = last_volume / cur_len
pre_3_price = float(obj[cur_len - 3][1])
if cur_price < close_dicts[stock]:
continue
price_chg_now = cur_price / pre_3_price
if price_chg_now < 1.011:
continue
#print("%d, %s %d %s cur_price: %f ,cur_volume: %f ,last_volume: %f ,sec_volume: %f ,avg_vol: %f ,last_avg_vol: %f, pre_3_price: %f" \
# %(idx, code, cur_len, cur_time, cur_price, cur_volume, last_volume, sec_volume, avg_vol, last_avg_vol, pre_3_price))
day_volume_ratio = avg_vol * 240 / volume_dicts[stock]
sec_volume_ratio = sec_volume / last_avg_vol
print("%d, %s %d %s price_up_chg: %f ,day_volume_ratio: %f ,sec_volume_ratio: %f " \
% (idx, code, cur_len, cur_time, price_chg_now, day_volume_ratio, sec_volume_ratio))
if price_chg_now < 1.04 and \
day_volume_ratio < 6 and \
sec_volume_ratio < 5 and \
(day_volume_ratio < 2 or sec_volume_ratio < 1.8):
continue
if not stock in bought:
buy_num = calc_buy_num(cur_price)
content = format('%s 模型 %s 买入 %s %d 股, 价格：%.2f' % (model, cur_time, stock, buy_num, cur_price))
if is_signal:
send_to_wx(content)
print(content)
buy_queries.append(build_table_buy_query(position_table, stock, cur_price, buy_num, day_volume_ratio, sec_volume_ratio,
price_chg_now))
bought.add(stock)
buy_dicts[stock] = int(cur_time)
count_dicts[stock] = total_len
except Exception as e:
print(e)
print('%s error' % stock)
time.sleep(10)
if len(buy_queries) > 0:
time.sleep(90 + idx * 5)
tcon = get_conn("db8")
execute_queries(tcon, buy_queries)
if __name__ == '__main__':
import sys
if not is_today_trading():
sys.exit("not trading day")
candidate_table = sys.argv[1]
model = sys.argv[2]
thread_count = int(sys.argv[3])
close_dicts = defaultdict(float)
volume_dicts = defaultdict(float)
is_signal = True
if len(sys.argv) >= 5:
is_signal = ('true' == sys.argv[4].lower())
stocks = set()
for table in [candidate_table]:
for stock in obtain_list_of_db_tickers(table):
stocks.add(stock)
print("total stocks len: %d" % len(stocks))
if len(stocks) == 0:
sys.exit("Empty in buy list")
for stock in stocks:
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20200616', end_date='20300318')
if df is None or len(df) < 1:
continue
close_dicts[stock] = float(df.head(1)['close'].max()) * 1.015
volume_dicts[stock] = float(df.head(1)['vol'].max())
print(close_dicts)
print(volume_dicts)
threads = []
for idx in range(0, thread_count):
thread = BuyThread(idx, thread_count, model, close_dicts, volume_dicts, is_signal)
thread.start()
threads.append(thread)
for t in threads:
t.join()
con.close()
print("Done")
Raw
logicbelief-master-puwork_buy_hk.py
import tushare as ts
import warnings
import threading
import time
print(ts.__version__)
warnings.simplefilter(action='ignore', category=FutureWarning)
import easyquotation
quotation = easyquotation.use("hkquote")
data = quotation.real(['00700'])
print(data)
# from send_wx import *
# from common import *
# from collections import defaultdict
#
# con = get_conn("db8")
#
# max_date = get_max_date()
# print(max_date)
#
# def obtain_list_of_db_tickers(table):
# cur = con.cursor()
# query_str = "select distinct code from %s where date = '%s'" % (table, max_date)
# cur.execute(query_str)
# data = cur.fetchall()
# cur.close()
# return [(d[0]) for d in data]
#
# class BuyThread(threading.Thread):
# def __init__(self, index, thread_count, model, close_dicts, volume_dicts, is_signal):
# threading.Thread.__init__(self)
# self.index = index
# self.thread_count = thread_count
# self.position_table = "%s_position" %(model)
# self.model = model
# self.close_dicts = close_dicts
# self.volume_dicts = volume_dicts
# self.is_signal = is_signal
#
# def run(self):
# print("Starting thread %d \n" % self.index)
# buy(self.index, self.thread_count, self.model, self.position_table, self.close_dicts, self.volume_dicts, self.is_signal)
#
# def buy(idx, thread_count, model, position_table, close_dicts, volume_dicts, is_signal):
# count_dicts = defaultdict(int)
# buy_dicts = defaultdict(int)
# buy_queries = []
# bought = set()
# while not is_after_hour():
# if not is_in_trade_time():
# time.sleep(60)
# continue
# i = -1
# for stock in stocks:
# i += 1
# if (i % thread_count) != idx:
# continue
# # print(stock)
# if stock in bought:
# continue
# code = stock[:-3]
# sc = stock[-2:].lower()
# try:
# key = "%s%s.js" % (sc, code)
# obj = quotation.real([code])[key]['time_data']
# total_len = len(obj)
# for cur_len in range(count_dicts[stock] - 1, total_len):
# if cur_len < 3:
# continue
# cur_time = obj[cur_len][0]
# if buy_dicts[stock] == int(cur_time):
# continue
# cur_price = float(obj[cur_len][1])
# cur_volume = float(obj[cur_len][2])
# last_volume = float(obj[cur_len - 1][2])
# sec_volume = cur_volume - last_volume
# avg_vol = cur_volume / (1 + cur_len)
# last_avg_vol = last_volume / cur_len
# pre_3_price = float(obj[cur_len - 3][1])
#
# if cur_price < close_dicts[stock]:
# continue
# price_chg_now = cur_price / pre_3_price
# if price_chg_now < 1.011:
# continue
# #print("%d, %s %d %s cur_price: %f ,cur_volume: %f ,last_volume: %f ,sec_volume: %f ,avg_vol: %f ,last_avg_vol: %f, pre_3_price: %f" \
# # %(idx, code, cur_len, cur_time, cur_price, cur_volume, last_volume, sec_volume, avg_vol, last_avg_vol, pre_3_price))
# day_volume_ratio = avg_vol * 240 / volume_dicts[stock]
# sec_volume_ratio = sec_volume / last_avg_vol
# print("%d, %s %d %s price_up_chg: %f ,day_volume_ratio: %f ,sec_volume_ratio: %f " \
# % (idx, code, cur_len, cur_time, price_chg_now, day_volume_ratio, sec_volume_ratio))
# if price_chg_now < 1.04 and \
# 6 > day_volume_ratio > 9 and \
# sec_volume_ratio < 5 and \
# (day_volume_ratio < 2 or sec_volume_ratio < 1.8):
# continue
# if not stock in bought:
# buy_num = calc_buy_num(cur_price)
# content = format('%s 模型 %s 买入 %s %d 股, 价格：%.2f' % (model, cur_time, stock, buy_num, cur_price))
# if is_signal:
# send_to_wx(content)
# print(content)
# buy_queries.append(build_table_buy_query(position_table, stock, cur_price, buy_num, day_volume_ratio, sec_volume_ratio,
# price_chg_now))
# bought.add(stock)
# buy_dicts[stock] = int(cur_time)
# count_dicts[stock] = total_len
# except Exception as e:
# print(e)
# print('%s error' % stock)
# time.sleep(10)
# if len(buy_queries) > 0:
# time.sleep(90 + idx * 5)
# tcon = get_conn("db8")
# execute_queries(tcon, buy_queries)
#
#
# if __name__ == '__main__':
# import sys
# candidate_table = sys.argv[1]
# model = sys.argv[2]
# thread_count = int(sys.argv[3])
# close_dicts = defaultdict(float)
# volume_dicts = defaultdict(float)
# is_signal = True
# if len(sys.argv) >= 5:
# is_signal = ('true' == sys.argv[4].lower())
#
# stocks = set()
# for table in [candidate_table]:
# for stock in obtain_list_of_db_tickers(table):
# stocks.add(stock)
#
# print("total stocks len: %d" % len(stocks))
# if len(stocks) == 0:
# sys.exit("Empty in buy list")
# for stock in stocks:
# df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20200316', end_date='20300318')
# if df is None or len(df) < 1:
# continue
# close_dicts[stock] = float(df.head(1)['close'].max()) * 1.015
# volume_dicts[stock] = float(df.head(1)['vol'].max())
#
# print(close_dicts)
# print(volume_dicts)
#
# threads = []
#
# for idx in range(0, thread_count):
# thread = BuyThread(idx, thread_count, model, close_dicts, volume_dicts, is_signal)
# thread.start()
# threads.append(thread)
#
# for t in threads:
# t.join()
# con.close()
# print("Done")
Raw
logicbelief-master-puwork_buy_us.py
import warnings
import threading
import os.path
import pandas as pd
import time
warnings.simplefilter(action='ignore', category=FutureWarning)
from send_wx import *
from common import *
from iex import *
con = get_conn()
max_date = get_max_date()
print(max_date)
def obtain_list_of_db_tickers(table):
cur = con.cursor()
query_str = "select distinct code from %s where date = '%s'" % (table, max_date)
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
from collections import defaultdict
close_dicts = defaultdict(float)
volume_dicts = defaultdict(float)
stocks = set()
for table in ['flat_us', 'reverse_us', 'vcpim_us']:
for stock in obtain_list_of_db_tickers(table):
stocks.add(stock)
print("total stocks len: %d" % len(stocks))
SOURCE_DIR = '/home/ruoang/ustocks/'
d = get_max_date()
candidates = []
for stock in stocks:
path = format("%s/%s.csv" %(SOURCE_DIR, stock))
if not os.path.isfile(path):
continue
try:
df = pd.read_csv(path)
if df is None or len(df) < 1:
continue
#print(df.tail(1))
close_dicts[stock] = float(df.tail(1)['close'].max())
volume_dicts[stock] = float(df.tail(1)['volume'].max())
except Exception as e:
print(e)
print('%s error' % stock)
print(close_dicts)
print(volume_dicts)
THREAD_COUNT = 1
class BuyThread(threading.Thread):
def __init__(self, index):
threading.Thread.__init__(self)
self.index = index
def run(self):
print("Starting thread %d \n" % self.index)
buy(self.index)
def buy(idx):
buy_dicts = defaultdict(int)
buy_queries = []
while is_us_trading_time():
i = -1
for stock in stocks:
i += 1
if (i % THREAD_COUNT) != idx:
continue
# print(stock)
try:
obj = get_intraday_prices(stock, 4)
total_len = len(obj)
if total_len < 4:
continue
#print(obj)
if obj[3]['close'] is None or obj[0]['close'] is None:
continue
cur_time = obj[3]['minute']
cur_price = float(obj[3]['close'])
pre_3_price = float(obj[0]['close'])
if buy_dicts[stock] == cur_time:
continue
print("%d, %s %s price_up_chg: %f" %(idx, stock, cur_time, cur_price/pre_3_price))
if cur_price < close_dicts[stock]:
continue
if cur_price / pre_3_price < 1.01:
continue
buy_num = calc_buy_us_num(cur_price)
content = format('%s 买入 %s %d 股, 价格: %.2f' % (cur_time, stock, buy_num, cur_price))
send_to_wx(content)
buy_queries.append(build_us_buy_query(stock, cur_price, buy_num, cur_price / pre_3_price))
buy_dicts[stock] = cur_time
except Exception as e:
print(e)
#print('%s error' % stock)
time.sleep(45)
print("Time is Out")
if len(buy_queries) > 0:
time.sleep(90 + idx * 5)
tcon = get_conn()
execute_queries(tcon, buy_queries)
threads = []
for idx in range(0, THREAD_COUNT):
thread = BuyThread(idx)
thread.start()
threads.append(thread)
for t in threads:
t.join()
con.close()
print("Done")
Raw
logicbelief-master-puwork_clean_st.py
from common import *
import warnings
warnings.simplefilter(action='ignore')
stocks = pro.stock_basic(exchange_id='', fields='ts_code, name, list_status')
to_delete_queries = []
for row in stocks.itertuples():
if 'ST' in row[2] or 'L' != row[3]:
update_str = "DELETE from stock_basic where stock = '%s'" % (row[1])
#print(update_str)
to_delete_queries.append(update_str)
#print('%s finish' % (row[1]))
con = get_conn("db8")
execute_queries(con, to_delete_queries)
con = get_conn("db")
execute_queries(con, to_delete_queries)
print("Done")
Raw
logicbelief-master-puwork_common.py
from datetime import datetime, timedelta
import tushare as ts
import warnings
import math
warnings.simplefilter(action='ignore', category=FutureWarning)
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
import MySQLdb as mdb
def is_in_trade_time():
now = datetime.now()
return is_trade_time(now)
def is_trade_time(now):
if now.hour < 9:
return False
if now.hour >= 15:
return False
if now.hour == 9 and now.minute < 30:
return False
if now.hour == 11 and now.minute > 30:
return False
if now.hour == 12:
return False
return True
def is_after_hour():
now = datetime.now()
return now.hour >= 15
def is_us_trading_time():
now = datetime.now()
return is_now_us_trading_time(now)
def is_now_us_trading_time(now):
return (now.hour == 21 and now.minute >= 30) or now.hour >= 22 or now.hour < 4
def get_max_date():
today = datetime.today()
return get_last_date(today)
def get_last_trade_date():
d = datetime.today()
today = d.strftime("%Y%m%d")
last = (d - timedelta(days=15)).strftime("%Y%m%d")
df = pro.query('trade_cal', start_date=last, end_date=today).query("is_open==1")
if d.weekday() < 5:
return df.tail(2)['cal_date'].min()
return df.tail(1)['cal_date'].min()
def get_last_date(d):
today = d.strftime("%Y%m%d")
last = (d - timedelta(days=15)).strftime("%Y%m%d")
df = pro.query('trade_cal', start_date=last, end_date=today).query("is_open==1")
min = df.tail(2)['cal_date'].min()
return datetime.strptime(min, "%Y%m%d").strftime("%Y-%m-%d")
def is_today_trading():
today = datetime.today().strftime("%Y%m%d")
df = pro.query('trade_cal', start_date=today, end_date=today).query("is_open==1")
return len(df) > 0
def execute_query(con, query):
queries = []
queries.append(query)
execute_queries(con, queries)
def execute_queries(con, queries):
try:
cur = con.cursor()
for query in queries:
print(query)
cur.execute(query)
cur.close()
con.commit()
con.close()
except Exception as e:
print(e)
def calc_buy_num(price, money=500):
num = math.floor(money / price) * 100
return num
def calc_buy_us_num(price, money=10000):
num = math.floor(money / price)
return num
def build_vcp_buy_query(stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct):
return "insert ignore vcp_position set code = '%s', buy_date=CURDATE(), price = '%s', volume = '%s', dvr = '%f', svr ='%f', up_pct = '%f' " % (stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct)
def build_buy_query(stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct):
return "insert ignore position set code = '%s', buy_date=CURDATE(), price = '%s', volume = '%s', dvr = '%f', svr ='%f', up_pct = '%f' " % (stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct)
def build_concept_buy_query(stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct):
return "insert ignore concept_position set code = '%s', buy_date=CURDATE(), price = '%s', volume = '%s', dvr = '%f', svr ='%f', up_pct = '%f' " % (stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct)
def build_table_buy_query(table, stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct):
return "insert ignore %s set code = '%s', buy_date=CURDATE(), price = '%s', volume = '%s', dvr = '%f', svr ='%f', up_pct = '%f' " % (table, stock, price, num, day_volume_ratio, sec_volume_ratio, up_pct)
def build_us_buy_query(stock, price, num, up_pct):
return "insert ignore us_position set code = '%s', buy_date=CURDATE(), price = '%s', volume = '%s', up_pct = '%f' " % (stock, price, num, up_pct)
def build_sell_query(position_table, now, volume, max_stop, stock, buy_date):
return "update %s set is_sold=1, sell_date=CURDATE(), sell_price = '%s', sell_volume = '%s', max_profit=%.2f where code='%s' and buy_date='%s' " % (
position_table, now, volume, max_stop * 100, stock, buy_date)
def obtain_stock_us(con):
cur = con.cursor()
query_str = "select stock from stock_us limit 0, 600"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
def obtain_stock_basic(con):
cur = con.cursor()
query_str = "select stock, name from stock_basic where total_mv > 500000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], d[1]) for d in data]
def obtain_stocks_above_mv(con, mv=500000):
cur = con.cursor()
query_str = "select stock, name from stock_basic where total_mv > %d" % mv
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], d[1]) for d in data]
class Stock:
def __init__(self, stock, days):
self.stock = stock
self.days = days
self.days_no = 0
def __str__(self):
return "stock: %s ,days: %s, days_no: %d" % (self.stock, self.days, self.days_no)
def get_conn(name="db"):
with open(name) as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
return con
def format_code(code):
code = code.strip()
if code.startswith("6"):
code += ".SH"
else:
code += ".SZ"
return code
if __name__ == '__main__':
print(is_in_trade_time())
print(is_trade_time(datetime.strptime("09:14:32", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("09:15:32", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("09:30:01", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("11:30:01", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("11:31:01", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("12:31:01", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("13:00:01", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("14:59:01", "%H:%M:%S")))
print(is_trade_time(datetime.strptime("15:00:01", "%H:%M:%S")))
print(is_after_hour())
print(get_max_date())
print(get_last_date(datetime.strptime("2020-03-23", "%Y-%m-%d")))
print(get_last_date(datetime.strptime("2020-03-24", "%Y-%m-%d")))
print(get_last_date(datetime.strptime("2020-03-22", "%Y-%m-%d")))
print(calc_buy_num(3.2))
print(calc_buy_num(5))
print(calc_buy_num(48))
print(build_buy_query("300000.SZ", 3.41, 5000, 1.5, 1.8, 1.235))
print(format("止盈, 目前盈利：%.2f ,最大盈利: %.2f" % (4.555, 5.2222)))
print(build_sell_query("hot", 10.2, 111000, 0.156, "000001.SZ", "2020-05-22"))
print(is_us_trading_time())
print(is_now_us_trading_time(datetime.strptime("23:32:01", "%H:%M:%S")))
print(is_now_us_trading_time(datetime.strptime("00:31:01", "%H:%M:%S")))
print(is_now_us_trading_time(datetime.strptime("01:31:01", "%H:%M:%S")))
print(is_now_us_trading_time(datetime.strptime("02:00:01", "%H:%M:%S")))
print(is_now_us_trading_time(datetime.strptime("03:59:01", "%H:%M:%S")))
print(is_now_us_trading_time(datetime.strptime("04:00:01", "%H:%M:%S")))
print(is_today_trading())
Raw
logicbelief-master-puwork_concept_buy.py
import tushare as ts
import warnings
import threading
import time
print(ts.__version__)
warnings.simplefilter(action='ignore', category=FutureWarning)
import easyquotation
quotation = easyquotation.use("timekline")
from common import *
con = get_conn("db8")
max_date = get_max_date()
print(max_date)
def obtain_list_of_db_tickers(table):
cur = con.cursor()
query_str = "select distinct code from %s where date = '%s'" % (table, max_date)
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
from collections import defaultdict
close_dicts = defaultdict(float)
volume_dicts = defaultdict(float)
stocks = set()
for table in ['top_concept']:
for stock in obtain_list_of_db_tickers(table):
stocks.add(stock)
print("total stocks len: %d" % len(stocks))
for stock in stocks:
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20200316', end_date='20300318')
if df is None or len(df) < 1:
continue
close_dicts[stock] = float(df.head(1)['close'].max()) * 1.015
volume_dicts[stock] = float(df.head(1)['vol'].max())
print(close_dicts)
print(volume_dicts)
THREAD_COUNT = 2
class BuyThread(threading.Thread):
def __init__(self, index):
threading.Thread.__init__(self)
self.index = index
def run(self):
print("Starting thread %d \n" % self.index)
buy(self.index)
def buy(idx):
count_dicts = defaultdict(int)
buy_dicts = defaultdict(int)
buy_queries = []
bought = set()
while not is_after_hour():
if not is_in_trade_time():
time.sleep(60)
continue
i = -1
for stock in stocks:
i += 1
if (i % THREAD_COUNT) != idx:
continue
# print(stock)
if stock in bought:
continue
code = stock[:-3]
sc = stock[-2:].lower()
try:
key = "%s%s.js" % (sc, code)
obj = quotation.real([code])[key]['time_data']
total_len = len(obj)
for cur_len in range(count_dicts[stock] - 1, total_len):
if cur_len < 3:
continue
cur_time = obj[cur_len][0]
if buy_dicts[stock] == int(cur_time):
continue
cur_price = float(obj[cur_len][1])
cur_volume = float(obj[cur_len][2])
last_volume = float(obj[cur_len - 1][2])
sec_volume = cur_volume - last_volume
avg_vol = cur_volume / (1 + cur_len)
last_avg_vol = last_volume / cur_len
pre_3_price = float(obj[cur_len - 3][1])
if cur_price < close_dicts[stock]:
continue
price_chg_now = cur_price / pre_3_price
if price_chg_now < 1.014 or price_chg_now >= 1.027:
continue
#print("%d, %s %d %s cur_price: %f ,cur_volume: %f ,last_volume: %f ,sec_volume: %f ,avg_vol: %f ,last_avg_vol: %f, pre_3_price: %f" \
# %(idx, code, cur_len, cur_time, cur_price, cur_volume, last_volume, sec_volume, avg_vol, last_avg_vol, pre_3_price))
day_volume_ratio = avg_vol * 240 / volume_dicts[stock]
sec_volume_ratio = sec_volume / last_avg_vol
print("%d, %s %d %s price_up_chg: %f ,day_volume_ratio: %f ,sec_volume_ratio: %f " \
% (idx, code, cur_len, cur_time, price_chg_now, day_volume_ratio, sec_volume_ratio))
if price_chg_now < 1.04 and \
6 > day_volume_ratio > 9 and \
sec_volume_ratio < 5 and \
(day_volume_ratio < 2 or sec_volume_ratio < 1.8):
continue
if not stock in bought:
buy_num = calc_buy_num(cur_price)
content = format('概念模型 %s 买入 %s %d 股, 价格：%.2f' % (cur_time, stock, buy_num, cur_price))
print(content)
buy_queries.append(build_concept_buy_query(stock, cur_price, buy_num, day_volume_ratio, sec_volume_ratio,
price_chg_now))
bought.add(stock)
buy_dicts[stock] = int(cur_time)
count_dicts[stock] = total_len
except Exception as e:
print(e)
print('%s error' % stock)
time.sleep(10)
if len(buy_queries) > 0:
time.sleep(90 + idx * 5)
tcon = get_conn("db8")
execute_queries(tcon, buy_queries)
if __name__ == '__main__':
import sys
if not is_today_trading():
sys.exit("not trading day")
threads = []
for idx in range(0, THREAD_COUNT):
thread = BuyThread(idx)
thread.start()
threads.append(thread)
for t in threads:
t.join()
con.close()
print("Done")
Raw
logicbelief-master-puwork_concept_sell.py
from datetime import date
import warnings
import easyquotation
import time
warnings.simplefilter(action='ignore')
from send_wx import *
from common import *
quotation = easyquotation.use('qq')
end_date = date.today()
end_date = end_date.strftime('%Y%m%d')
con = get_conn("db8")
def obtain_position():
cur = con.cursor()
query_str = "select code, buy_date, price, volume from concept_position where is_sold = 0"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], d[1], d[2], d[3]) for d in data]
from collections import defaultdict
high_dicts = defaultdict(float)
positions = obtain_position()
for (stock, buy_date, price, volume) in positions:
start_date = (buy_date + timedelta(days=1)).strftime('%Y%m%d')
end_date = date.today() + timedelta(days=-1)
end_date = end_date.strftime('%Y%m%d')
try:
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date=start_date, end_date=end_date)
if df is None or len(df) < 1:
continue
# print(df)
high_dicts[stock] = float(df['high'].max())
except Exception as e:
print(e)
print('obtain %s error' % (stock))
sold = set()
sell_queries = []
while not is_after_hour():
if not is_in_trade_time():
time.sleep(60)
continue
for (stock, buy_date, price, volume) in positions:
if stock in sold:
continue
# print("%s %s %s %s %s" %(stock, buy_date, price, volume, predict))
code = stock[0:6]
try:
obj = quotation.real(code)
now = obj[code]["now"]
stop = (now - price) / price
is_sell = False
msg = ""
# print(stop)
if stop <= -0.03:
is_sell = True
msg = format("概念模型止损, 亏损：%.2f, 价格: %.2f" % (stop * 100, now))
elif stop > 0:
high = obj[code]["high"]
today_max_stop = (high - price) / price
if high_dicts[stock] > high:
high = high_dicts[stock]
max_stop = (high - price) / price
if max_stop > 0.2:
if stop <= 0.8 * max_stop:
is_sell = True
elif max_stop > 0.1:
if max_stop - stop > 0.05:
is_sell = True
elif max_stop > 0.06:
if max_stop - stop > 0.03:
is_sell = True
elif today_max_stop - stop >= 0.03:
is_sell = True
if is_sell:
msg = format("概念模型止盈, 目前盈利：%.2f ,最大盈利: %.2f, 价格: %.2f" % (stop * 100, max_stop * 100, now))
if is_sell:
content = format("%s: %s" % (code, msg))
#send_to_wx(content)
print(content)
sold.add(stock)
sell_query = "update concept_position set is_sold=1, sell_date=CURDATE(), sell_price = '%s', sell_volume = '%s' where code='%s' " % (
now, volume, stock)
sell_queries.append(sell_query)
except Exception as e:
print(e)
print('%s error' % (stock))
time.sleep(30)
con.close()
con = get_conn("db8")
execute_queries(con, sell_queries)
print("Done")
Raw
logicbelief-master-puwork_cross_buy.py
import tushare as ts
import warnings
import threading
import time
import sys
print(ts.__version__)
warnings.simplefilter(action='ignore', category=FutureWarning)
import easyquotation
quotation = easyquotation.use("timekline")
from send_wx import *
from common import *
from collections import defaultdict
con = get_conn("db8")
max_date = get_max_date()
print(max_date)
def obtain_list_of_db_tickers(table):
cur = con.cursor()
query_str = "select distinct code from %s where date = '%s'" % (table, max_date)
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
class BuyThread(threading.Thread):
def __init__(self, index, thread_count, model, close_dicts, volume_dicts, is_signal):
threading.Thread.__init__(self)
self.index = index
self.thread_count = thread_count
self.position_table = "%s_position" %(model)
self.model = model
self.close_dicts = close_dicts
self.volume_dicts = volume_dicts
self.is_signal = is_signal
def run(self):
print("Starting thread %d \n" % self.index)
buy(self.index, self.thread_count, self.model, self.position_table, self.close_dicts, self.volume_dicts, self.is_signal)
def buy(idx, thread_count, model, position_table, close_dicts, volume_dicts, is_signal):
count_dicts = defaultdict(int)
buy_dicts = defaultdict(int)
buy_queries = []
bought = set()
while not is_after_hour():
if not is_in_trade_time():
time.sleep(60)
continue
i = -1
for stock in stocks:
i += 1
if (i % thread_count) != idx:
continue
# print(stock)
if stock in bought:
continue
code = stock[:-3]
sc = stock[-2:].lower()
try:
key = "%s%s.js" % (sc, code)
obj = quotation.real([code])[key]['time_data']
total_len = len(obj)
for cur_len in range(count_dicts[stock] - 1, total_len):
if cur_len < 3:
continue
cur_time = obj[cur_len][0]
if buy_dicts[stock] == int(cur_time):
continue
cur_price = float(obj[cur_len][1])
cur_volume = float(obj[cur_len][2])
last_volume = float(obj[cur_len - 1][2])
sec_volume = cur_volume - last_volume
avg_vol = cur_volume / (1 + cur_len)
last_avg_vol = last_volume / cur_len
pre_3_price = float(obj[cur_len - 3][1])
if cur_price < close_dicts[stock]:
continue
price_chg_now = cur_price / pre_3_price
if price_chg_now < 1.02:
continue
#print("%d, %s %d %s cur_price: %f ,cur_volume: %f ,last_volume: %f ,sec_volume: %f ,avg_vol: %f ,last_avg_vol: %f, pre_3_price: %f" \
# %(idx, code, cur_len, cur_time, cur_price, cur_volume, last_volume, sec_volume, avg_vol, last_avg_vol, pre_3_price))
day_volume_ratio = avg_vol * 240 / volume_dicts[stock]
sec_volume_ratio = sec_volume / last_avg_vol
print("%d, %s %d %s price_up_chg: %f ,day_volume_ratio: %f ,sec_volume_ratio: %f " \
% (idx, code, cur_len, cur_time, price_chg_now, day_volume_ratio, sec_volume_ratio))
if price_chg_now < 1.04 and \
6 > day_volume_ratio > 9 and \
sec_volume_ratio < 5 and \
(day_volume_ratio < 2 or sec_volume_ratio < 1.8):
continue
if not stock in bought:
buy_num = calc_buy_num(cur_price)
content = format('%s 模型 %s 买入 %s %d 股, 价格：%.2f' % (model, cur_time, stock, buy_num, cur_price))
if is_signal:
send_to_wx(content)
print(content)
buy_queries.append(build_table_buy_query(position_table, stock, cur_price, buy_num, day_volume_ratio, sec_volume_ratio,
price_chg_now))
bought.add(stock)
buy_dicts[stock] = int(cur_time)
count_dicts[stock] = total_len
except Exception as e:
print(e)
print('%s error' % stock)
time.sleep(10)
if len(buy_queries) > 0:
time.sleep(90 + idx * 5)
tcon = get_conn("db8")
execute_queries(tcon, buy_queries)
if __name__ == '__main__':
import sys
if not is_today_trading():
sys.exit("not trading day")
candidate_table = sys.argv[1]
model = sys.argv[2]
thread_count = int(sys.argv[3])
close_dicts = defaultdict(float)
volume_dicts = defaultdict(float)
is_signal = True
if len(sys.argv) >= 5:
is_signal = ('true' == sys.argv[4].lower())
stocks = set()
for table in [candidate_table]:
for stock in obtain_list_of_db_tickers(table):
stocks.add(stock)
print("total stocks len: %d" % len(stocks))
if len(stocks) == 0:
sys.exit("Empty in buy list")
for stock in stocks:
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20200316', end_date='20300318')
if df is None or len(df) < 1:
continue
close_dicts[stock] = float(df.head(1)['close'].max()) * 1.015
volume_dicts[stock] = float(df.head(1)['vol'].max())
print(close_dicts)
print(volume_dicts)
threads = []
for idx in range(0, thread_count):
thread = BuyThread(idx, thread_count, model, close_dicts, volume_dicts, is_signal)
thread.start()
threads.append(thread)
for t in threads:
t.join()
con.close()
print("Done")
Raw
logicbelief-master-puwork_cross_local.py
import os.path
import pandas as pd
from common import *
warnings.simplefilter(action='ignore', category=FutureWarning)
def obtain_stock_basic_test():
return [("603713.SH", "")]
def days_above(df, close):
if df is None:
return 0
days = 0
for idx in df.index:
high = df['high'][idx]
if close < high:
break
days = days + 1
return days
def ma_days_above(df, ma='ma5', limit=5):
if df is None:
return False
days = 0
last = 100000000
for idx in df.index:
cur = df[ma][idx]
if cur < last:
days = days + 1
if days >= limit:
return True
last = cur
#print(days)
return days >= limit
def is_cross(open, close, high, low):
if abs(open/close - 1) > 0.008:
return False
return (high/low - 1) < 0.06
count = 0
date = 0
con = get_conn("db8")
stocks = obtain_stock_basic(con)
con.close()
stock_len = len(stocks)
from collections import defaultdict
dict = defaultdict(list)
SOURCE_DIR = '/home/ruoang/astocks/'
for (stock, name) in stocks:
if stock.startswith('688'):
continue
#print(stock)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
path = format("%s/%s.csv" %(SOURCE_DIR, stock))
if not os.path.isfile(path):
continue
try:
dfo = pd.read_csv(path, header=0, index_col=0)
#dfo = dfo.set_index('trade_date')
#dfo = dfo[::-1]
except Exception as e:
print(e)
print('%s error' % stock)
continue
for i in range(0, 1):
df = dfo[i:]
last = df.head(1)
#print(last)
if last is None:
continue
# if last['amount'].max() < 100000:
# continue
last_close = last['close'].max()
# if last_close < 6:
# continue
last_high = last['high'].max()
last_low = last['low'].max()
last_open = last['open'].max()
last_vol = last['vol'].max()
last_150 = last['ma150'].max()
if last_150 is None:
continue
last_50 = last['ma50'].max()
last_5 = last['ma5'].max()
lastv_5 = last['ma_v_5'].max()
last_pct_chg = last['pct_chg'].max()
if last_pct_chg > 4 or last_5 < last_50:
continue
if not is_cross(last_open, last_close, last_high, last_low):
continue
if not ma_days_above(df.head(7)):
continue
if last_vol/lastv_5 > 0.6:
continue
d = int(last['trade_date'])
remain = df[1:]
dict[d].append(Stock(stock, 0))
print('%s %s' %(stock, name))
con = get_conn("db8")
for (k, v) in dict.items():
sorted_days = sorted(v, key=lambda item: item.days)
lens = len(v)
n = 0
for s in reversed(sorted_days):
if n >= 64:
break
#print(k)
print(str(s))
column_str = """date, code, value"""
final_str = "INSERT IGNORE INTO top_cross (%s) VALUES (%d, '%s', %d)" % \
(column_str, k, s.stock, s.days)
n += 1
cur = con.cursor()
print(final_str)
cur.execute(final_str)
con.commit()
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_daily_profit_loss.py
import pandas as pd
from datetime import date
from common import *
print(ts.__version__)
warnings.simplefilter(action='ignore', category=FutureWarning)
def position_profit():
cur = con.cursor()
query_str = "select code, profit, profit_pct, is_sold, sell_date from position"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], d[1], d[2], d[3], d[4]) for d in data]
date_today = date.today().strftime('%Y-%m-%d')
con = get_conn()
# cur = con.cursor()
# query_str = "select * from position"
# cur.execute(query_str)
# des = cur.description
# print(des)
# cur.close()
position_pro = position_profit()
df = pd.DataFrame(position_pro, columns = ['stock', 'profit', 'profit_pct', 'is_sold', 'sell_date'])
print(df.shape[0])
# print(df[(df['is_sold'] == 1) & (df['sell_date'] == date_today)])
daily_profit = df[df['is_sold'] == 0].sum()['profit'] +\
df[(df['is_sold'] == 1) & (df['sell_date'] == date_today)].sum()['profit']
daily_profit_pct = df[df['is_sold'] == 0].sum()['profit_pct'] +\
df[(df['is_sold'] == 1) & (df['sell_date'] == date_today)].sum()['profit_pct']
win_cnt = df[(df['is_sold'] == 0) & (df['profit'] > 0)].shape[0] +\
df[(df['is_sold'] == 1) & (df['sell_date'] == date_today) & (df['profit'] > 0)].shape[0]
loss_cnt = df[(df['is_sold'] == 0) & (df['profit'] < 0)].shape[0] +\
df[(df['is_sold'] == 1) & (df['sell_date'] == date_today) & (df['profit'] < 0)].shape[0]
print(daily_profit)
print(daily_profit_pct)
print(win_cnt)
print(loss_cnt)
con.close()
print("Done")
Raw
logicbelief-master-puwork_download_cn.py
from common import *
warnings.simplefilter(action='ignore')
con = get_conn("db8")
def obtain_stock_basic():
cur = con.cursor()
query_str = "select distinct stock from stock_basic where total_mv > 100000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
start = (datetime.today() - timedelta(days=700)).strftime("%Y%m%d")
print(start)
end = datetime(2030, 1, 1).strftime("%Y%m%d")
for stock in obtain_stock_basic():
try:
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date=start, end_date=end, ma=[5, 20, 50, 150, 200])
df = df.head(300)
df.to_csv("/home/ruoang/astocks/" + stock + '.csv')
print('%s finish' % (stock))
except Exception as e:
print(e)
print('%s error' % (stock))
con.close()
print("Done")
Raw
logicbelief-master-puwork_download_hk.py
from common import *
warnings.simplefilter(action='ignore')
con = get_conn("db8")
def obtain_stock_basic():
cur = con.cursor()
query_str = "select distinct stock from stock_basic where total_mv > 100000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
start = (datetime.today() - timedelta(days=500)).strftime("%Y%m%d")
print(start)
end = datetime(2030, 1, 1).strftime("%Y%m%d")
basic = pro.hk_basic()
for index, row in basic.iterrows():
stock = row['ts_code']
try:
df = pro.hk_daily(ts_code=stock, start_date=start, end_date=end, ma=[5, 20, 50, 150, 200])
df = df.head(300)
df.to_csv("/home/ruoang/hstocks/" + stock + '.csv')
print('%s finish' % (stock))
except Exception as e:
print(e)
print('%s error' % (stock))
con.close()
print("Done")
Raw
logicbelief-master-puwork_download_new_us.py
import pandas_datareader.data as web
from common import *
import os
warnings.simplefilter(action='ignore')
con = get_conn()
start = datetime(2017, 1, 1)
print(start)
end = datetime(2030, 1, 1)
SOURCE_DIR = '/home/ruoang/ustocks/'
for stock in obtain_stock_us(con):
try:
path = format("%s/%s.csv" % (SOURCE_DIR, stock))
if os.path.isfile(path):
continue
df = web.DataReader(stock, 'iex', start, end, access_key='pk_88d98ab1d0344ceb8184b898313a18cc')
df.to_csv("/home/ruoang/ustocks/" + stock + '.csv', mode='a', header=True)
print('%s finish' % (stock))
except Exception as e:
print(e)
print('%s error' % (stock))
con.close()
print("Done")
Raw
logicbelief-master-puwork_download_us.py
import pandas_datareader.data as web
from common import *
warnings.simplefilter(action='ignore')
con = get_conn()
yesterday = datetime.today() - timedelta(days=1)
start = datetime(2020, yesterday.month, yesterday.day)
print(start)
end = datetime(2030, 1, 1)
for stock in obtain_stock_us(con):
try:
df = web.DataReader(stock, 'iex', start, end, access_key='pk_88d98ab1d0344ceb8184b898313a18cc')
df.to_csv("/home/ruoang/ustocks/" + stock + '.csv', mode='a', header=False)
print('%s finish' % (stock))
except Exception as e:
print(e)
print('%s error' % (stock))
con.close()
print("Done")
Raw
logicbelief-master-puwork_factor.py
import math
import pandas as pd
def days_above(df, ma):
df = df.iloc[::-1]
if df is None:
return 0
days = 0
diff = df[ma] - df[ma].shift(1)
#print(diff)
for idx in reversed(diff.index):
#print(idx)
if math.isnan(diff[idx]) or diff[idx] < 0:
continue
days = days + 1
# print(days)
return days
def up_days(df):
if df is None:
return 0
return len(df[df.pct_chg > 0.0])
def close_up_days(df):
if df is None:
return 0
days = 0
for idx in range(0, len(df)):
row = df.iloc[idx]
if row['close'] > (row['high'] + row['low']) / 2:
days = days + 1
# print(days)
return days
def up_down_v_ratio(df):
if df is None:
return 0
up_volume = 0.0
down_volume = 1.0
for idx in range(0, len(df)):
row = df.iloc[idx]
if row['pct_chg'] >= 0.0:
up_volume += row['vol']
else:
down_volume += row['vol']
return up_volume / down_volume
def up_down_v_ratio_v2(df):
if df is None:
return 0
up_volume = 0.0
down_volume = 1.0
up_cnt = 0
down_cnt = 0
for idx in range(0, len(df)):
row = df.iloc[idx]
if row['pct_chg'] >= 0.0:
up_volume += row['vol']
up_cnt += 1
else:
down_volume += row['vol']
down_cnt += 1
if up_cnt == 0:
return 0
if down_cnt == 0 or down_volume < 100:
return 10000
return (up_volume/ up_cnt) / (down_volume/down_cnt)
def up_down_volume_ratio_v2(df):
if df is None:
return 0
up_volume = 0.0
down_volume = 1.0
up_cnt = 0
down_cnt = 0
for idx in range(0, len(df)):
row = df.iloc[idx]
if row['pct_chg'] >= 0.0:
up_volume += row['volume']
up_cnt += 1
else:
down_volume += row['volume']
down_cnt += 1
if up_cnt == 0:
return 0
if down_cnt == 0 or down_volume < 100:
return 10000
return (up_volume/ up_cnt) / (down_volume/down_cnt)
def up_down_pct_chg_ratio(df):
if df is None:
return 0
up_price_chg = 0.0
down_price_chg = 0.0
up_cnt = 0
down_cnt = 0
for idx in range(0, len(df)):
row = df.iloc[idx]
if row['pct_chg'] >= 0.0:
up_price_chg += row['pct_chg']
up_cnt += 1
else:
down_price_chg -= row['pct_chg']
down_cnt += 1
if up_cnt == 0:
return 0
if down_cnt == 0 or down_price_chg < 0.1:
return 1000
return (up_price_chg/ up_cnt) / (down_price_chg/down_cnt)
def avg_volatility(df):
if df is None:
return 0
volatility = 0.0
for idx in range(0, len(df)):
row = df.iloc[idx]
volatility += (row['high'] / row['low']) - 1
return volatility * 100 / len(df)
def avg_pct_chg(df):
return df['pct_chg'].mean()
def days_small_above_large(df, ma_small, ma_large):
df = df.iloc[::-1]
if df is None:
return 0
days = 0
diff = df[ma_small] - df[ma_large]
#print(diff)
diff_diff = diff - diff.shift(1)
#print(diff_diff)
for idx in reversed(diff_diff.index):
#print(idx)
if math.isnan(diff_diff[idx]) or diff_diff[idx] < 0 or diff[idx] < 0:
continue
days = days + 1
#print(days)
return days
def max_drawback(df):
if df is None:
return 0
df['max2here'] = df['close'].expanding().max()
df['dd2here'] = df['close']/df['max2here']
end_date, remains = tuple(df.sort_values(by=['dd2here']).iloc[0]
[['trade_date', 'dd2here']])
# start_date = df[df['trade_date'] <= end_date]\
# .sort_values(by='close', ascending=False)\
# .iloc[0]['trade_date']
# print(start_date)
# print(end_date)
return (round((1-remains) * 100, 2))
def max_profit(df):
col_names = ['trade_date']
maxseq = seq = pd.DataFrame(columns=col_names)
start, end, sum_start = -1, -1, -1
maxsum_, sum_ = 0, 0
i = 0
for x in df.itertuples():
if math.isnan(x[9]):
continue
seq.loc[len(seq)] = x[2]; sum_ += x[9]
if maxsum_ < sum_:
maxseq = seq; maxsum_ = sum_
start, end = sum_start, i
elif sum_ < 0:
seq = pd.DataFrame(columns=col_names); sum_ = 0
sum_start = i
i += 1
#assert maxsum_ == sum(maxseq[:end - start])
#print(maxsum_)
#print(maxseq)
return round(maxsum_, 2)
def ma(data, n=10, val_name="close"):
import numpy as np
'''
移动平均线 Moving Average
Parameters
------
data:pandas.DataFrame
通过 get_h_data 取得的股票数据
n:int
移动平均线时长，时间单位根据data决定
val_name:string
计算哪一列的列名，默认为 close 收盘值
return
-------
list
移动平均线
'''
values = []
MA = []
for index, row in data.iterrows():
values.append(row[val_name])
if len(values) == n:
del values[0]
MA.append(np.average(values))
return np.asarray(MA)
def sma(data, column, n=10, val_name="close"):
ma = pd.Series(data[val_name].rolling(window=n).mean(), name = column)
data = data.join(ma)
return data
Raw
logicbelief-master-puwork_first_zt.py
from common import *
import MySQLdb as mdb
import math
print(math.isnan(1))
with open("db8") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock, name from stock_basic where total_mv > 200000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], d[1]) for d in data]
def obtain_stock_basic_test():
return [("601158.SH", "")]
def flat_days_above(df):
if df is None:
return False
return range_high_chg(df.head(5), 0.35)
def range_close_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['close'].max() / df['close'].min() - 1 < ratio
def range_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['high'].max() / df['low'].min() - 1 < ratio
class Stock:
def __init__(self, stock, days):
self.stock = stock
self.days = days
self.days_no = 0
def __str__(self):
return "stock: %s ,days: %s, days_no: %d" % (self.stock, self.days, self.days_no)
count = 0
date = 0
stocks = obtain_stock_basic()
con.close()
stock_len = len(stocks)
from collections import defaultdict
dict = defaultdict(list)
for (stock, name) in stocks:
#print(stock)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
dfo = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20200101', end_date='20220710', ma=[5, 10])
if dfo is None:
continue
for i in range(10, 100):
df = dfo[i:]
last = df.head(1)
if last is None:
continue
last_pct_chg = last['pct_chg'].max()
if last_pct_chg < 9.9:
continue
last_vol = last['vol'].max()
if last_vol is None:
continue
lastv_10 = last['ma_v_10'].max()
if lastv_10 is None:
continue
# if last_close < last_5 or last_5 < last_50 or last_50 < last_150:
# continue
if last['trade_date'].empty:
continue
#if not flat_days_above(df):
# continue
d = int(last['trade_date'])
remain = df[1:]
if remain.head(5)['pct_chg'].abs().sum() > 10:
continue
before_5_vol = remain.head(5)['vol'].mean()
if before_5_vol is None:
continue
if last_vol < 2 * before_5_vol:
continue
if last_vol > 3 * before_5_vol:
continue
dict[d].append(Stock(stock, last_vol / before_5_vol / remain.head(5)['pct_chg'].abs().mean()))
# print('%s %s' %(stock, name))
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
for (k, v) in dict.items():
sorted_days = sorted(v, key=lambda item: item.days)
lens = len(v)
n = 0
for s in reversed(sorted_days):
if n >= 3:
break
if s is None or s.days is None:
continue
if math.isnan(s.days):
continue
print(k)
print(str(s))
column_str = """date, code, value"""
final_str = "INSERT IGNORE INTO first_zt5 (%s) VALUES (%d, '%s', %d)" % \
(column_str, k, s.stock, s.days)
n += 1
cur = con.cursor()
cur.execute(final_str)
con.commit()
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_flat_backtest.py
from common import *
warnings.simplefilter(action='ignore', category=FutureWarning)
def days_above(df, close):
if df is None:
return 0
days = 0
for idx in df.index:
high = df['high'][idx]
if close < high:
break
days = days + 1
return days
def ma_days_above(df, ma='ma150', limit=7):
if df is None:
return False
days = 0
last = 100000000
for idx in df.index:
cur = df[ma][idx]
if cur > last:
break
days = days + 1
if days >= limit:
return True
last = cur
#print(days)
return days >= limit
def flat_days_above(df):
if df is None:
return False
if range_high_chg(df.head(3), 0.02) or \
range_close_chg(df.head(3), 0.025) or \
range_high_chg(df.head(5), 0.07):
return True
return False
def range_close_chg(df, ratio):
print(df['close'].max() / df['close'].min())
return df['close'].max() / df['close'].min() - 1 < ratio
def range_high_chg(df, ratio):
print(df)
print(df['high'].max() / df['low'].min())
return df['high'].max() / df['low'].min() - 1 < ratio
if __name__ == '__main__':
import sys
stock = format_code(sys.argv[1])
end_date = sys.argv[2]
if stock.startswith('688'):
sys.exit("not support 688")
#print(stock)
try:
dfo = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20190101', end_date=end_date, ma=[5, 50, 150])
except Exception as e:
print(e)
print('%s error' % stock)
sys.exit("read data error")
for i in range(0, 1):
df = dfo[i:]
last = df.head(1)
#print(last)
if last is None:
continue
if last['amount'].max() < 100000:
continue
last_close = last['close'].max()
if last_close < 6:
continue
last_high = last['high'].max()
last_open = last['open'].max()
last_vol = last['vol'].max()
last_150 = last['ma150'].max()
if last_150 is None:
continue
last_50 = last['ma50'].max()
last_5 = last['ma5'].max()
last_pct_chg = last['pct_chg'].max()
if last_pct_chg > 4 or last_5 < last_50:
continue
if not ma_days_above(df):
continue
if not flat_days_above(df):
continue
d = int(last['trade_date'])
remain = df[1:]
days = days_above(remain, last_high)
if days >= len(df) - 1:
continue
days = days_above(remain, last_close * 1.03)
#print(days)
if days < 20:
continue
print('%s hit' %(stock))
Raw
logicbelief-master-puwork_flat_local.py
import os.path
import pandas as pd
from common import *
warnings.simplefilter(action='ignore', category=FutureWarning)
def obtain_stock_basic_test():
return [("603713.SH", "")]
def days_above(df, close):
if df is None:
return 0
days = 0
for idx in df.index:
high = df['high'][idx]
if close < high:
break
days = days + 1
return days
def ma_days_above(df, ma='ma150', limit=7):
if df is None:
return False
days = 0
last = 100000000
for idx in df.index:
cur = df[ma][idx]
if cur > last:
break
days = days + 1
if days >= limit:
return True
last = cur
#print(days)
return days >= limit
def flat_days_above(df):
if df is None:
return False
if range_high_chg(df.head(3), 0.02) or \
range_high_chg(df.head(5), 0.07):
return True
return False
def range_close_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['close'].max() / df['close'].min() - 1 < ratio
def range_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['high'].max() / df['low'].min() - 1 < ratio
def vcp_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
idxmax = df['high'].idxmax()
idxmin = df['low'].idxmin()
if df.iloc[idxmax]['trade_date'] > df.iloc[idxmin]['trade_date']:
return False
return df['high'].max() / df['low'].min() - 1 < ratio
count = 0
date = 0
con = get_conn()
stocks = obtain_stock_basic(con)
con.close()
stock_len = len(stocks)
from collections import defaultdict
dict = defaultdict(list)
SOURCE_DIR = '/home/ruoang/astocks/'
for (stock, name) in stocks:
if stock.startswith('688'):
continue
#print(stock)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
path = format("%s/%s.csv" %(SOURCE_DIR, stock))
if not os.path.isfile(path):
continue
try:
dfo = pd.read_csv(path, header=0, index_col=0)
#dfo = dfo.set_index('trade_date')
#dfo = dfo[::-1]
except Exception as e:
print(e)
print('%s error' % stock)
continue
for i in range(0, 1):
df = dfo[i:]
last = df.head(1)
#print(last)
if last is None:
continue
if last['amount'].max() < 100000:
continue
last_close = last['close'].max()
if last_close < 6:
continue
last_high = last['high'].max()
last_open = last['open'].max()
last_vol = last['vol'].max()
last_150 = last['ma150'].max()
if last_150 is None:
continue
last_50 = last['ma50'].max()
last_5 = last['ma5'].max()
last_pct_chg = last['pct_chg'].max()
if last_pct_chg > 4 or last_5 < last_50 or last_50 < last_150:
continue
if not ma_days_above(df):
continue
if not flat_days_above(df):
continue
d = int(last['trade_date'])
remain = df[1:]
days = days_above(remain, last_high)
if days >= len(df) - 1:
continue
days = days_above(remain, last_close * 1.03)
#print(days)
if days < 20:
continue
dict[d].append(Stock(stock, days))
print('%s %s' %(stock, name))
con = get_conn()
for (k, v) in dict.items():
sorted_days = sorted(v, key=lambda item: item.days)
lens = len(v)
n = 0
for s in reversed(sorted_days):
if n >= 64:
break
#print(k)
print(str(s))
column_str = """date, code, value"""
final_str = "INSERT IGNORE INTO flat_nb (%s) VALUES (%d, '%s', %d)" % \
(column_str, k, s.stock, s.days)
n += 1
cur = con.cursor()
print(final_str)
cur.execute(final_str)
con.commit()
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_flat_local_close.py
import os.path
import pandas as pd
from common import *
warnings.simplefilter(action='ignore', category=FutureWarning)
def obtain_stock_basic_test():
return [("603713.SH", "")]
def days_above(df, close):
if df is None:
return 0
days = 0
for idx in df.index:
high = df['close'][idx]
if close < high:
break
days = days + 1
return days
def ma_days_above(df, ma='ma150', limit=7):
if df is None:
return False
days = 0
last = 100000000
for idx in df.index:
cur = df[ma][idx]
if cur > last:
break
days = days + 1
if days >= limit:
return True
last = cur
#print(days)
return days >= limit
def flat_days_above(df):
if df is None:
return False
if range_high_chg(df.head(3), 0.02) or \
range_high_chg(df.head(5), 0.07):
return True
return False
def range_close_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['close'].max() / df['close'].min() - 1 < ratio
def range_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['high'].max() / df['low'].min() - 1 < ratio
def vcp_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
idxmax = df['high'].idxmax()
idxmin = df['low'].idxmin()
if df.iloc[idxmax]['trade_date'] > df.iloc[idxmin]['trade_date']:
return False
return df['high'].max() / df['low'].min() - 1 < ratio
count = 0
date = 0
con = get_conn()
stocks = obtain_stock_basic(con)
con.close()
stock_len = len(stocks)
from collections import defaultdict
dict = defaultdict(list)
SOURCE_DIR = '/home/ruoang/astocks/'
for (stock, name) in stocks:
if stock.startswith('688'):
continue
#print(stock)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
path = format("%s/%s.csv" %(SOURCE_DIR, stock))
if not os.path.isfile(path):
continue
try:
dfo = pd.read_csv(path, header=0, index_col=0)
#dfo = dfo.set_index('trade_date')
#dfo = dfo[::-1]
except Exception as e:
print(e)
print('%s error' % stock)
continue
for i in range(0, 1):
df = dfo[i:]
last = df.head(1)
#print(last)
if last is None:
continue
if last['amount'].max() < 100000:
continue
last_close = last['close'].max()
if last_close < 6:
continue
last_high = last['high'].max()
last_open = last['open'].max()
last_vol = last['vol'].max()
last_150 = last['ma150'].max()
if last_150 is None:
continue
last_50 = last['ma50'].max()
last_5 = last['ma5'].max()
last_pct_chg = last['pct_chg'].max()
if last_pct_chg > 4 or last_5 < last_50 or last_50 < last_150:
continue
if not ma_days_above(df):
continue
if not flat_days_above(df):
continue
d = int(last['trade_date'])
remain = df[1:]
days = days_above(remain, last_high)
if days >= len(df) - 1:
continue
days = days_above(remain, last_close * 1.03)
#print(days)
if days < 20:
continue
dict[d].append(Stock(stock, days))
print('%s %s' %(stock, name))
con = get_conn("db8")
for (k, v) in dict.items():
sorted_days = sorted(v, key=lambda item: item.days)
lens = len(v)
n = 0
for s in reversed(sorted_days):
if n >= 64:
break
#print(k)
print(str(s))
column_str = """date, code, value"""
final_str = "INSERT IGNORE INTO flat_close (%s) VALUES (%d, '%s', %d)" % \
(column_str, k, s.stock, s.days)
n += 1
cur = con.cursor()
print(final_str)
cur.execute(final_str)
con.commit()
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_flat_nb.py
from common import *
warnings.simplefilter(action='ignore', category=FutureWarning)
import MySQLdb as mdb
con = get_conn()
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock, name from stock_basic where total_mv > 500000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], d[1]) for d in data]
def obtain_stock_basic_test():
return [("300206.SZ", "")]
def days_above(df, close):
if df is None:
return 0
days = 0
for idx in df.index:
high = df['high'][idx]
if close < high:
break
days = days + 1
return days
def ma_days_above(df, ma='ma150', limit=7):
if df is None:
return False
days = 0
last = 100000000
for idx in df.index:
cur = df[ma][idx]
if cur > last:
break
days = days + 1
if days >= limit:
return True
last = cur
#print(days)
return days >= limit
def flat_days_above(df):
if df is None:
return False
if range_high_chg(df.head(5), 0.08) or \
range_high_chg(df.head(10), 0.12):
return True
return vcp_high_chg(df.head(7), 0.10) and \
vcp_high_chg(df.head(15), 0.20) or \
vcp_high_chg(df.head(30), 0.30)
def range_close_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['close'].max() / df['close'].min() - 1 < ratio
def range_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['high'].max() / df['low'].min() - 1 < ratio
def vcp_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
idxmax = df['high'].idxmax()
idxmin = df['low'].idxmin()
if df.iloc[idxmax]['trade_date'] > df.iloc[idxmin]['trade_date']:
return False
return df['high'].max() / df['low'].min() - 1 < ratio
class Stock:
def __init__(self, stock, days):
self.stock = stock
self.days = days
self.days_no = 0
def __str__(self):
return "stock: %s ,days: %s, days_no: %d" % (self.stock, self.days, self.days_no)
count = 0
date = 0
stocks = obtain_stock_basic_test()
con.close()
stock_len = len(stocks)
from collections import defaultdict
dict = defaultdict(list)
for (stock, name) in stocks:
#print(stock)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
dfo = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20190101', end_date='20200609', ma=[5, 50, 150])
if dfo is None:
continue
for i in range(0, 1):
df = dfo[i:]
last = df.head(1)
if last is None:
continue
if last['amount'].max() < 100000:
continue
last_close = last['close'].max()
last_open = last['open'].max()
last_vol = last['vol'].max()
last_150 = last['ma150'].max()
if last_150 is None:
continue
last_50 = last['ma50'].max()
last_5 = last['ma5'].max()
lastv_50 = last['ma_v_50'].max()
last_pct_chg = last['pct_chg'].max()
if last_5 < last_50 or last_50 < last_150:
continue
if last['trade_date'].empty:
continue
# if not ma_days_above(df):
# continue
if not flat_days_above(df):
continue
d = int(last['trade_date'])
remain = df[1:]
days = days_above(remain, last_close * 1.04)
#print(days)
if days < 80:
continue
dict[d].append(Stock(stock, days))
# print('%s %s' %(stock, name))
con = get_conn()
for (k, v) in dict.items():
sorted_days = sorted(v, key=lambda item: item.days)
lens = len(v)
n = 0
for s in reversed(sorted_days):
if n >= 64:
break
#print(k)
print(str(s))
column_str = """date, code, value"""
final_str = "INSERT IGNORE INTO flat_test (%s) VALUES (%d, '%s', %d)" % \
(column_str, k, s.stock, s.days)
n += 1
cur = con.cursor()
#cur.execute(final_str)
con.commit()
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_flat_us.py
from common import *
from factor import *
import os.path
import pandas as pd
warnings.simplefilter(action='ignore', category=FutureWarning)
con = get_conn()
def obtain_stock_basic_test():
return [("601158.SH", "")]
def days_above(df, close):
if df is None:
return 0
days = 0
for idx in reversed(df.index):
high = df['high'][idx]
if close < high:
break
days = days + 1
return days
def ma_days_above(df, ma='ma150', limit=7):
if df is None:
return False
days = 0
last = 0
for idx in df.index:
cur = df[ma][idx]
if cur < last:
break
days = days + 1
if days >= limit:
return True
last = cur
#print(days)
return days >= limit
def flat_days_above(df):
if df is None:
return False
return range_close_chg(df.tail(10), 0.04) or \
range_close_chg(df.tail(5), 0.02) or \
range_close_chg(df.tail(20), 0.05) or \
range_high_chg(df.tail(10), 0.06) or \
range_high_chg(df.tail(20), 0.08)
def range_close_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['close'].max() / df['close'].min() - 1 < ratio
def range_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['high'].max() / df['low'].min() - 1 < ratio
class Stock:
def __init__(self, stock, days):
self.stock = stock
self.days = days
self.days_no = 0
def __str__(self):
return "stock: %s ,days: %s, days_no: %d" % (self.stock, self.days, self.days_no)
count = 0
date = 0
stocks = obtain_stock_us(con)
con.close()
stock_len = len(stocks)
from collections import defaultdict
dict = defaultdict(list)
SOURCE_DIR = '/home/ruoang/ustocks/'
yesterday = datetime.today() - timedelta(days=1)
d = datetime(2020, yesterday.month, yesterday.day).strftime("%Y-%m-%d")
print(d)
candidates = []
for stock in stocks:
#print(stock)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
path = format("%s/%s.csv" %(SOURCE_DIR, stock))
if not os.path.isfile(path):
continue
try:
df = pd.read_csv(path)
df = sma(df, "ma5", n=5)
df = sma(df, "ma50", n=50)
df = sma(df, "ma150", n=150)
except Exception as e:
#print(e)
#print('%s error' % stock)
continue
last = df.tail(1)
if last is None:
continue
if last['date'].max() != d:
print("Exception %s wrong date %s" % (stock, last['date'].max()))
continue
#print(last)
last_close = last['close'].max()
last_open = last['open'].max()
last_vol = last['volume'].max()
last_150 = last['ma150'].max()
last_50 = last['ma50'].max()
last_5 = last['ma5'].max()
if last_5 < last_50 or last_50 < last_150:
continue
if not ma_days_above(df.tail(7)):
continue
if not flat_days_above(df):
continue
remain = df[-200:-1]
days = days_above(remain, last_close * 1.03)
if days < 40:
continue
candidates.append(Stock(stock, days))
con = get_conn()
sorted_days = sorted(candidates, key=lambda item: item.days)
n = 0
for s in reversed(sorted_days):
if n >= 10:
break
print(str(s))
column_str = """date, code, value"""
final_str = "INSERT IGNORE INTO flat_us (%s) VALUES ('%s', '%s', %d)" % \
(column_str, d, s.stock, s.days)
n += 1
cur = con.cursor()
cur.execute(final_str)
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ggt_local.py
from common import *
from collections import defaultdict
pct_chg_dict = defaultdict(float)
vol_dict = defaultdict(float)
candidates = set()
def obtain_stock_basic():
# 获取沪股通成分
df = pro.hs_const(hs_type='SH')
for index, row in df.iterrows():
candidates.add(row['ts_code'])
# 获取深股通成分
df = pro.hs_const(hs_type='SZ')
for index, row in df.iterrows():
candidates.add(row['ts_code'])
today = datetime.today().strftime("%Y%m%d")
today_daily = pro.daily(trade_date=today)
for index, row in today_daily.iterrows():
stock = row['ts_code']
if not(stock in candidates):
continue
if float(row['pct_chg']) < 1:
continue
pct_chg_dict[stock] = float(row['pct_chg'])
vol_dict[stock] = float(row['amount'])
sorted_concepts = sorted(pct_chg_dict.items(), key=lambda d: d[1])
print("Done")
Raw
logicbelief-master-puwork_git.sh
git status
git add .
git commit -m "update"
git push
Raw
logicbelief-master-puwork_hot_local.py
import os.path
import pandas as pd
from common import *
warnings.simplefilter(action='ignore', category=FutureWarning)
def obtain_stock_basic_test():
return [("603713.SH", "")]
def days_above(df, close):
if df is None:
return 0
days = 0
for idx in df.index:
high = df['high'][idx]
if close < high:
break
days = days + 1
return days
def ma_days_above(df, ma='ma150', limit=7):
if df is None:
return False
days = 0
last = 100000000
for idx in df.index:
cur = df[ma][idx]
if cur > last:
break
days = days + 1
if days >= limit:
return True
last = cur
#print(days)
return days >= limit
def flat_days_above(df):
if df is None:
return False
if range_high_chg(df.head(3), 0.05) or \
range_high_chg(df.head(5), 0.09) or \
range_high_chg(df.head(10), 0.15):
return True
return False
def range_close_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['close'].max() / df['close'].min() - 1 < ratio
def range_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
return df['high'].max() / df['low'].min() - 1 < ratio
def vcp_high_chg(df, ratio):
#print(df['close'].max() / df['close'].min())
idxmax = df['high'].idxmax()
idxmin = df['low'].idxmin()
if df.iloc[idxmax]['trade_date'] > df.iloc[idxmin]['trade_date']:
return False
return df['high'].max() / df['low'].min() - 1 < ratio
count = 0
date = 0
con = get_conn("db8")
max_date = datetime.today().strftime("%Y-%m-%d")
print(max_date)
def obtain_list_of_db_tickers(table):
cur = con.cursor()
query_str = "select distinct code from %s where date = '%s'" % (table, max_date)
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
stocks = set()
for table in ['top_concept', 'top_industry']:
for stock in obtain_list_of_db_tickers(table):
stocks.add(stock)
print("total stocks len: %d" % len(stocks))
con.close()
stock_len = len(stocks)
from collections import defaultdict
dict = defaultdict(list)
SOURCE_DIR = '/home/ruoang/astocks/'
for stock in stocks:
#print(stock)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
path = format("%s/%s.csv" %(SOURCE_DIR, stock))
if not os.path.isfile(path):
continue
try:
dfo = pd.read_csv(path, header=0, index_col=0)
#dfo = dfo.set_index('trade_date')
#dfo = dfo[::-1]
except Exception as e:
print(e)
print('%s error' % stock)
continue
for i in range(0, 1):
df = dfo[i:]
last = df.head(1)
#print(last)
if last is None:
continue
if last['amount'].max() < 100000:
continue
last_close = last['close'].max()
last_high = last['high'].max()
last_open = last['open'].max()
last_vol = last['vol'].max()
last_150 = last['ma150'].max()
if last_150 is None:
continue
last_50 = last['ma50'].max()
last_5 = last['ma5'].max()
lastv_50 = last['ma_v_50'].max()
last_pct_chg = last['pct_chg'].max()
if last_pct_chg > 3:
continue
if not flat_days_above(df):
continue
d = int(last['trade_date'])
remain = df[1:]
days = days_above(remain, last_close * 1.03)
#print(days)
if days < 20:
continue
dict[d].append(Stock(stock, days))
print('%s' %(stock))
con = get_conn("db8")
for (k, v) in dict.items():
sorted_days = sorted(v, key=lambda item: item.days)
lens = len(v)
n = 0
for s in reversed(sorted_days):
if n >= 64:
break
#print(k)
print(str(s))
column_str = """date, code, value"""
final_str = "INSERT IGNORE INTO hot (%s) VALUES (%d, '%s', %d)" % \
(column_str, k, s.stock, s.days)
n += 1
cur = con.cursor()
print(final_str)
cur.execute(final_str)
con.commit()
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_hurst.py
import time
import pandas as pd
from numpy import cumsum, log, polyfit, sqrt, std, subtract
from numpy.random import randn
# Import the Time Series library
import statsmodels.tsa.stattools as tts
import warnings
warnings.simplefilter(action='ignore')
def hurst(ts):
"""Returns the Hurst Exponent of the time series vector ts"""
# Create the range of lag values
lags = range(2, 100)
# Calculate the array of the variances of the lagged differences
tau = [sqrt(std(subtract(ts[lag:], ts[:-lag]))) for lag in lags]
# Use a linear fit to estimate the Hurst Exponent
poly = polyfit(log(lags), log(tau), 1)
# Return the Hurst exponent from the polyfit output
return poly[0] * 2.0
def is_trend(df):
try:
adf = tts.adfuller(df['close'], 1)
if adf[0] < adf[4]['5%']:
#print("adf(%s): %s" % (stock, adf))
hurstValue = hurst(df['close'])
if hurstValue < 0.5:
return True
except ValueError:
pass
return False
Raw
logicbelief-master-puwork_iex.py
import json
import requests
URL_PREFIX = 'https://cloud.iexapis.com/stable/stock/'
URL_SUFFIX = '/intraday-prices?token=pk_88d98ab1d0344ceb8184b898313a18cc&chartIEXOnly=true&chartLast='
def get_intraday_prices(stock, n=390):
url = URL_PREFIX + stock + URL_SUFFIX + str(n)
r = requests.get(url)
s = str(r.content, encoding = "utf-8")
return json.loads(s)
if __name__ == '__main__':
json = get_intraday_prices('AMZN', 3)
print(len(json))
for item in json:
print(item)
print(item['volume'])
print(item['close'])
Raw
logicbelief-master-puwork_income.py
from common import *
from collections import defaultdict
def get_ebit_dict(period):
last_income = pro.income_vip(period = str(period), fields='ts_code, ebit')
ebit_dict = defaultdict(float)
for index, cash in last_income.iterrows():
code = cash['ts_code']
ebit = float(cash['ebit'])
if math.isnan(ebit) or ebit is None or ebit < 50000000:
continue
ebit_dict[code] = ebit
income = pro.income_vip(period = str(period + 10000), fields='ts_code, ebit')
ebit_pct_dict = defaultdict(float)
for index, cash in income.iterrows():
code = cash['ts_code']
if not (code in ebit_dict):
continue
ebit = float(cash['ebit'])
if ebit / ebit_dict[code] > 1.1:
ebit_pct_dict[code] = ebit / ebit_dict[code]
#print("%s %.2f" %(code, ebit / ebit_dict[code]))
return ebit_pct_dict
if __name__ == '__main__':
cur_ebit_pct = get_ebit_dict(20190331)
last_ebit_pct = get_ebit_dict(20181231)
queries = []
for (k, v) in cur_ebit_pct.items():
if not (k in last_ebit_pct):
continue
if v > last_ebit_pct[k]:
print("%s %.2f %.2f" %(k, v, last_ebit_pct[k]))
query = "INSERT IGNORE INTO watch (code) VALUES ('%s')" % (k)
queries.append(query)
con = get_conn("db8")
execute_queries(con, queries)
print("Done")
Raw
logicbelief-master-puwork_industries.txt
煤炭开采
环境保护
食品
水泥
广告包装
批发业
公路
旅游服务
机械基件
服饰
矿物制品
机场
元器件
生物制药
水力发电
超市连锁
农药化肥
钢加工
橡胶
饲料
白酒
新型电力
中成药
IT设备
多元金融
铝
商品城
渔业
啤酒
摩托车
仓储物流
汽车配件
日用化工
铅锌
水务
焦炭加工
运输设备
None
小金属
银行
其他商业
家用电器
黄金
酒店餐饮
出版业
汽车整车
装修装饰
区域地产
专用机械
证券
化工原料
染料涂料
陶瓷
港口
林业
百货
农业综合
石油贸易
其他建材
轻工机械
汽车服务
软饮料
电信运营
乳制品
农用机械
旅游景点
造纸
家居用品
房产服务
综合类
石油开采
工程机械
铜
通信设备
纺织机械
电器仪表
软件服务
保险
园区开发
机床制造
路桥
建筑工程
医疗保健
供气供热
船舶
电气设备
化工机械
文教休闲
玻璃
互联网
电器连锁
火力发电
半导体
红黄酒
影视音像
种植业
全国地产
普钢
公共交通
纺织
铁路
特种钢
石油加工
化纤
医药商业
空运
塑料
航空
水运
化学制药
商贸代理
Raw
logicbelief-master-puwork_industry_buy.py
import tushare as ts
import warnings
import threading
import time
import sys
print(ts.__version__)
warnings.simplefilter(action='ignore', category=FutureWarning)
import easyquotation
quotation = easyquotation.use("timekline")
from send_wx import *
from common import *
from collections import defaultdict
con = get_conn("db8")
max_date = get_max_date()
print(max_date)
def obtain_list_of_db_tickers(table):
cur = con.cursor()
query_str = "select distinct code from %s where date = '%s'" % (table, max_date)
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
class BuyThread(threading.Thread):
def __init__(self, index, thread_count, model, close_dicts, volume_dicts, is_signal):
threading.Thread.__init__(self)
self.index = index
self.thread_count = thread_count
self.position_table = "%s_position" %(model)
self.model = model
self.close_dicts = close_dicts
self.volume_dicts = volume_dicts
self.is_signal = is_signal
def run(self):
print("Starting thread %d \n" % self.index)
buy(self.index, self.thread_count, self.model, self.position_table, self.close_dicts, self.volume_dicts, self.is_signal)
def buy(idx, thread_count, model, position_table, close_dicts, volume_dicts, is_signal):
count_dicts = defaultdict(int)
buy_dicts = defaultdict(int)
buy_queries = []
bought = set()
while not is_after_hour():
if not is_in_trade_time():
time.sleep(60)
continue
i = -1
for stock in stocks:
i += 1
if (i % thread_count) != idx:
continue
# print(stock)
if stock in bought:
continue
code = stock[:-3]
sc = stock[-2:].lower()
try:
key = "%s%s.js" % (sc, code)
obj = quotation.real([code])[key]['time_data']
total_len = len(obj)
for cur_len in range(count_dicts[stock] - 1, total_len):
if cur_len < 3:
continue
cur_time = obj[cur_len][0]
if buy_dicts[stock] == int(cur_time):
continue
cur_price = float(obj[cur_len][1])
cur_volume = float(obj[cur_len][2])
last_volume = float(obj[cur_len - 1][2])
sec_volume = cur_volume - last_volume
avg_vol = cur_volume / (1 + cur_len)
last_avg_vol = last_volume / cur_len
pre_3_price = float(obj[cur_len - 3][1])
if cur_price < close_dicts[stock]:
continue
price_chg_now = cur_price / pre_3_price
if price_chg_now < 1.011:
continue
#print("%d, %s %d %s cur_price: %f ,cur_volume: %f ,last_volume: %f ,sec_volume: %f ,avg_vol: %f ,last_avg_vol: %f, pre_3_price: %f" \
# %(idx, code, cur_len, cur_time, cur_price, cur_volume, last_volume, sec_volume, avg_vol, last_avg_vol, pre_3_price))
day_volume_ratio = avg_vol * 240 / volume_dicts[stock]
sec_volume_ratio = sec_volume / last_avg_vol
print("%d, %s %d %s price_up_chg: %f ,day_volume_ratio: %f ,sec_volume_ratio: %f " \
% (idx, code, cur_len, cur_time, price_chg_now, day_volume_ratio, sec_volume_ratio))
if price_chg_now < 1.04 and \
day_volume_ratio <= 8.1 and \
sec_volume_ratio < 5 and \
(day_volume_ratio < 2 or sec_volume_ratio < 1.8):
continue
if not stock in bought:
buy_num = calc_buy_num(cur_price)
content = format('%s 模型 %s 买入 %s %d 股, 价格：%.2f' % (model, cur_time, stock, buy_num, cur_price))
if is_signal:
send_to_wx(content)
print(content)
buy_queries.append(build_table_buy_query(position_table, stock, cur_price, buy_num, day_volume_ratio, sec_volume_ratio,
price_chg_now))
bought.add(stock)
buy_dicts[stock] = int(cur_time)
count_dicts[stock] = total_len
except Exception as e:
print(e)
print('%s error' % stock)
time.sleep(10)
if len(buy_queries) > 0:
time.sleep(90 + idx * 5)
tcon = get_conn("db8")
execute_queries(tcon, buy_queries)
if __name__ == '__main__':
import sys
if not is_today_trading():
sys.exit("not trading day")
candidate_table = sys.argv[1]
model = sys.argv[2]
thread_count = int(sys.argv[3])
close_dicts = defaultdict(float)
volume_dicts = defaultdict(float)
is_signal = True
if len(sys.argv) >= 5:
is_signal = ('true' == sys.argv[4].lower())
stocks = set()
for table in [candidate_table]:
for stock in obtain_list_of_db_tickers(table):
stocks.add(stock)
print("total stocks len: %d" % len(stocks))
if len(stocks) == 0:
sys.exit("Empty in buy list")
for stock in stocks:
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date='20200316', end_date='20300318')
if df is None or len(df) < 1:
continue
close_dicts[stock] = float(df.head(1)['close'].max()) * 1.015
volume_dicts[stock] = float(df.head(1)['vol'].max())
print(close_dicts)
print(volume_dicts)
threads = []
for idx in range(0, thread_count):
thread = BuyThread(idx, thread_count, model, close_dicts, volume_dicts, is_signal)
thread.start()
threads.append(thread)
for t in threads:
t.join()
con.close()
print("Done")
Raw
logicbelief-master-puwork_leader.py
from common import *
import warnings
import MySQLdb as mdb
warnings.simplefilter(action='ignore', category=FutureWarning)
data = pro.stock_basic(exchange_id='', list_status='L', fields='ts_code,name,list_date')
stock_len = len(data)
count = 0
start_date = (datetime.today() - timedelta(days=7)).strftime("%Y%m%d")
end_date = datetime.today().strftime("%Y%m%d")
queries = []
for row in data.itertuples():
listdate = int(row[3])
if listdate > 20191001:
continue
stock = row[1]
name = row[2]
code = stock.split('.', 1)
count += 1
if count % 100 == 0:
print('finish %d/%d' % (count, stock_len))
try:
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date=start_date, end_date=end_date)
except Exception as e:
continue
pct_chg = df.head(1)['close'].max() / df.tail(1)['open'].max() - 1
if pct_chg > 0.35:
print('%s %s %.2f' %(stock, name, pct_chg))
column_str = """date, code, name, pct_chg"""
final_str = "INSERT IGNORE INTO leader (%s) VALUES ('%s', '%s', '%s', %.2f)" % \
(column_str, end_date, stock, name, pct_chg)
queries.append(final_str)
con = get_conn("db8")
execute_queries(con, queries)
print("Done")
Raw
logicbelief-master-puwork_leading.py
from common import *
from collections import defaultdict
pct_chg_dict = defaultdict(float)
amp_dict = defaultdict(float)
vol_ratio_dict = defaultdict(float)
candidates = set()
con = get_conn("db8")
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_basic where total_mv > 300000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
for stock in obtain_stock_basic():
candidates.add(stock)
today = datetime.today().strftime("%Y%m%d")
today_daily = pro.daily(trade_date=today)
for index, row in today_daily.iterrows():
stock = row['ts_code']
pct_chg_dict[stock] = float(row['pct_chg'])
amp_dict[stock] = (float(row['high']) / float(row['low']) - 1) * 100
sorted_pct_chg = sorted(pct_chg_dict.items(), key=lambda d: d[1])
sorted_amp = sorted(amp_dict.items(), key=lambda d: d[1])
today_daily_basic = pro.daily_basic(trade_date=today)
for index, row in today_daily_basic.iterrows():
stock = row['ts_code']
if math.isnan(row['volume_ratio']):
continue
vol_ratio_dict[stock] = float(row['volume_ratio'])
sorted_pct_chg = sorted(pct_chg_dict.items(), key=lambda d: d[1])
sorted_amp = sorted(amp_dict.items(), key=lambda d: d[1])
sorted_vol_ratio = sorted(vol_ratio_dict.items(), key=lambda d: d[1])
n = 70
a = set(map(lambda t: t[0], sorted_pct_chg[-n:]))
b = set(map(lambda t: t[0], sorted_amp[-n:]))
c = set(map(lambda t: t[0], sorted_vol_ratio[-n:]))
queries = []
column_str = """date, code, value"""
today = datetime.today().strftime("%Y-%m-%d")
for stock in a.intersection(b).intersection(c).intersection(candidates):
final_str = "INSERT IGNORE INTO top_leading (%s) VALUES ('%s', '%s', %d)" % \
(column_str, today, stock, 0)
queries.append(final_str)
execute_queries(con, queries)
print("Done")
Raw
logicbelief-master-puwork_log_watcher.py
import os
import glob2
from send_wx import *
for path in glob2.glob("/home/ruoang/*.out"):
print(path)
if not os.path.isfile(path):
continue
with open(path) as f:
content = f.readlines()
hasException = ""
for line in content:
if "exception" in line or "Traceback" in line or "can't" in line:
hasException = True
break
if hasException:
name = path.split('/')[3]
send_to_wx("%s 有异常，请及时处理" % (name))
Raw
logicbelief-master-puwork_loss_pct_distribution.py
from common import *
start_date = (datetime.today() - timedelta(days=8)).strftime("%Y%m%d")
end_date = (datetime.today() + timedelta(days=1)).strftime("%Y%m%d")
today = (datetime.today()).strftime("%Y-%m-%d")
profit_date = (datetime.today() - timedelta(days=1)).strftime("%Y-%m-%d")
def obtain_position(con, table):
cur = con.cursor()
query_str = "select code, buy_date, profit_pct from %s where is_sold = 1 and profit > 0" %(table)
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], d[1], d[2]) for d in data]
con = get_conn()
for table in ['position']:
for (stock, buy_date, profit_pct) in obtain_position(con, table):
pre_date = (buy_date - timedelta(days=1)).strftime("%Y%m%d")
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date=pre_date, end_date=pre_date)
if df is None or len(df) < 1:
continue
print("%s, %s, %.2f, %.2f" %(stock, buy_date, df.head(1)['pct_chg'].max(), profit_pct))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_lstm_appendix_02_01_python_versions.py
# check library version numbers
# scipy
import scipy
print('scipy: %s' % scipy.__version__)
# numpy
import numpy
print('numpy: %s' % numpy.__version__)
# matplotlib
import matplotlib
print('matplotlib: %s' % matplotlib.__version__)
# pandas
import pandas
print('pandas: %s' % pandas.__version__)
# statsmodels
import statsmodels
print('statsmodels: %s' % statsmodels.__version__)
# scikit-learn
import sklearn
print('sklearn: %s' % sklearn.__version__)
Raw
logicbelief-master-puwork_lstm_appendix_02_02_deep_versions.py
# theano
import theano
print('theano: %s' % theano.__version__)
# tensorflow
import tensorflow
print('tensorflow: %s' % tensorflow.__version__)
# keras
import keras
print('keras: %s' % keras.__version__)
Raw
logicbelief-master-puwork_lstm_chapter_06_01_time_series_to_supervised.py
# transform univariate time series to supervised learning problem
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define univariate time series
series = array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
print(series.shape)
# transform to a supervised learning problem
X, y = split_sequence(series, 3)
print(X.shape, y.shape)
# show each sample
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_06_02_transform_univariate_2d_3d.py
# transform univariate 2d to 3d
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define univariate time series
series = array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
print(series.shape)
# transform to a supervised learning problem
X, y = split_sequence(series, 3)
print(X.shape, y.shape)
# transform input from [samples, features] to [samples, timesteps, features]
X = X.reshape((X.shape[0], X.shape[1], 1))
print(X.shape)
Raw
logicbelief-master-puwork_lstm_chapter_06_03_example_load_data.py
# example of defining a dataset
from numpy import array
# define the dataset
data = list()
n = 5000
for i in range(n):
data.append([i+1, (i+1)*10])
data = array(data)
print(data[:5, :])
print(data.shape)
Raw
logicbelief-master-puwork_lstm_chapter_06_04_example_drop_time.py
# example of dropping the time dimension from the dataset
from numpy import array
# define the dataset
data = list()
n = 5000
for i in range(n):
data.append([i+1, (i+1)*10])
data = array(data)
# drop time
data = data[:, 1]
print(data.shape)
Raw
logicbelief-master-puwork_lstm_chapter_06_05_example_split_subsequences.py
# example of splitting a univariate sequence into subsequences
from numpy import array
# define the dataset
data = list()
n = 5000
for i in range(n):
data.append([i+1, (i+1)*10])
data = array(data)
# drop time
data = data[:, 1]
# split into samples (e.g. 5000/200 = 25)
samples = list()
length = 200
# step over the 5,000 in jumps of 200
for i in range(0,n,length):
# grab from i to i + 200
sample = data[i:i+length]
samples.append(sample)
print(len(samples))
Raw
logicbelief-master-puwork_lstm_chapter_06_06_example_create_array.py
# example of creating an array of subsequence
from numpy import array
# define the dataset
data = list()
n = 5000
for i in range(n):
data.append([i+1, (i+1)*10])
data = array(data)
# drop time
data = data[:, 1]
# split into samples (e.g. 5000/200 = 25)
samples = list()
length = 200
# step over the 5,000 in jumps of 200
for i in range(0,n,length):
# grab from i to i + 200
sample = data[i:i+length]
samples.append(sample)
# convert list of arrays into 2d array
data = array(samples)
print(data.shape)
Raw
logicbelief-master-puwork_lstm_chapter_06_07_example_reshape_3d.py
# example of creating a 3d array of subsequences
from numpy import array
# define the dataset
data = list()
n = 5000
for i in range(n):
data.append([i+1, (i+1)*10])
data = array(data)
# drop time
data = data[:, 1]
# split into samples (e.g. 5000/200 = 25)
samples = list()
length = 200
# step over the 5,000 in jumps of 200
for i in range(0,n,length):
# grab from i to i + 200
sample = data[i:i+length]
samples.append(sample)
# convert list of arrays into 2d array
data = array(samples)
# reshape into [samples, timesteps, features]
data = data.reshape((len(samples), length, 1))
print(data.shape)
Raw
logicbelief-master-puwork_lstm_chapter_07_01_univariate_dataset.py
# univariate data preparation
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_07_02_mlp_univariate.py
# univariate mlp example
from numpy import array
from keras.models import Sequential
from keras.layers import Dense
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# define model
model = Sequential()
model.add(Dense(100, activation='relu', input_dim=n_steps))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_07_03_dependent_time_series.py
# multivariate data preparation
from numpy import array
from numpy import hstack
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
print(dataset)
Raw
logicbelief-master-puwork_lstm_chapter_07_04_transform_dependent_time_series.py
# multivariate data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_07_05_mlp_dependent_time_series.py
# multivariate mlp example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# flatten input
n_input = X.shape[1] * X.shape[2]
X = X.reshape((X.shape[0], n_input))
# define model
model = Sequential()
model.add(Dense(100, activation='relu', input_dim=n_input))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[80, 85], [90, 95], [100, 105]])
x_input = x_input.reshape((1, n_input))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_07_06_multiheaded_mlp_dependent_time_series.py
# multivariate mlp example
from numpy import array
from numpy import hstack
from keras.models import Model
from keras.layers import Input
from keras.layers import Dense
from keras.layers.merge import concatenate
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# separate input data
X1 = X[:, :, 0]
X2 = X[:, :, 1]
# first input model
visible1 = Input(shape=(n_steps,))
dense1 = Dense(100, activation='relu')(visible1)
# second input model
visible2 = Input(shape=(n_steps,))
dense2 = Dense(100, activation='relu')(visible2)
# merge input models
merge = concatenate([dense1, dense2])
output = Dense(1)(merge)
model = Model(inputs=[visible1, visible2], outputs=output)
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit([X1, X2], y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[80, 85], [90, 95], [100, 105]])
x1 = x_input[:, 0].reshape((1, n_steps))
x2 = x_input[:, 1].reshape((1, n_steps))
yhat = model.predict([x1, x2], verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_07_07_split_multivariate_time_series.py
# multivariate output data prep
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_07_08_mlp_multivariate_time_series.py
# multivariate output mlp example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# flatten input
n_input = X.shape[1] * X.shape[2]
X = X.reshape((X.shape[0], n_input))
n_output = y.shape[1]
# define model
model = Sequential()
model.add(Dense(100, activation='relu', input_dim=n_input))
model.add(Dense(n_output))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[70,75,145], [80,85,165], [90,95,185]])
x_input = x_input.reshape((1, n_input))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_07_09_multi-output_mlp_multivariate_time_series.py
# multivariate output mlp example
from numpy import array
from numpy import hstack
from keras.models import Model
from keras.layers import Input
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# flatten input
n_input = X.shape[1] * X.shape[2]
X = X.reshape((X.shape[0], n_input))
# separate output
y1 = y[:, 0].reshape((y.shape[0], 1))
y2 = y[:, 1].reshape((y.shape[0], 1))
y3 = y[:, 2].reshape((y.shape[0], 1))
# define model
visible = Input(shape=(n_input,))
dense = Dense(100, activation='relu')(visible)
# define output 1
output1 = Dense(1)(dense)
# define output 2
output2 = Dense(1)(dense)
# define output 2
output3 = Dense(1)(dense)
# tie together
model = Model(inputs=visible, outputs=[output1, output2, output3])
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, [y1,y2,y3], epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[70,75,145], [80,85,165], [90,95,185]])
x_input = x_input.reshape((1, n_input))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_07_10_data_prep_multi_step_forecasting.py
# multi-step data preparation
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the sequence
if out_end_ix > len(sequence):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix:out_end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# split into samples
X, y = split_sequence(raw_seq, n_steps_in, n_steps_out)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_07_11_mlp_multi_step_forecast.py
# univariate multi-step vector-output mlp example
from numpy import array
from keras.models import Sequential
from keras.layers import Dense
# split a univariate sequence into samples
def split_sequence(sequence, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the sequence
if out_end_ix > len(sequence):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix:out_end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# split into samples
X, y = split_sequence(raw_seq, n_steps_in, n_steps_out)
# define model
model = Sequential()
model.add(Dense(100, activation='relu', input_dim=n_steps_in))
model.add(Dense(n_steps_out))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps_in))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_07_12_prepare_data_multi_step_dependent_series.py
# multivariate multi-step data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out-1
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1:out_end_ix, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_07_13_mlp_multi_step_dependent_series.py
# multivariate multi-step mlp example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out-1
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1:out_end_ix, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
# flatten input
n_input = X.shape[1] * X.shape[2]
X = X.reshape((X.shape[0], n_input))
# define model
model = Sequential()
model.add(Dense(100, activation='relu', input_dim=n_input))
model.add(Dense(n_steps_out))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[70, 75], [80, 85], [90, 95]])
x_input = x_input.reshape((1, n_input))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_07_14_prepare_data_multi_step_multivariate_series.py
# multivariate multi-step data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix:out_end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_07_15_mlp_multi_step_multivariate_series.py
# multivariate multi-step mlp example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix:out_end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
# flatten input
n_input = X.shape[1] * X.shape[2]
X = X.reshape((X.shape[0], n_input))
# flatten output
n_output = y.shape[1] * y.shape[2]
y = y.reshape((y.shape[0], n_output))
# define model
model = Sequential()
model.add(Dense(100, activation='relu', input_dim=n_input))
model.add(Dense(n_output))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[60, 65, 125], [70, 75, 145], [80, 85, 165]])
x_input = x_input.reshape((1, n_input))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_01_univariate_dataset.py
# univariate data preparation
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_08_02_cnn_univariate.py
# univariate cnn example
from numpy import array
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# reshape from [samples, timesteps] into [samples, timesteps, features]
n_features = 1
X = X.reshape((X.shape[0], X.shape[1], n_features))
# define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=2, activation='relu', input_shape=(n_steps, n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(50, activation='relu'))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=1000, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_03_dependent_time_series_dataset.py
# multivariate data preparation
from numpy import array
from numpy import hstack
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
print(dataset)
Raw
logicbelief-master-puwork_lstm_chapter_08_04_split_samples_dependent_time_series.py
# multivariate data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_08_05_cnn_multivariate_dependent_series.py
# multivariate cnn example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=2, activation='relu', input_shape=(n_steps, n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(50, activation='relu'))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=1000, verbose=0)
# demonstrate prediction
x_input = array([[80, 85], [90, 95], [100, 105]])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_06_multiheaded_cnn_multivariate_dependent_series.py
# multivariate multi-headed 1d cnn example
from numpy import array
from numpy import hstack
from keras.models import Model
from keras.layers import Input
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.layers.merge import concatenate
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# one time series per head
n_features = 1
# separate input data
X1 = X[:, :, 0].reshape(X.shape[0], X.shape[1], n_features)
X2 = X[:, :, 1].reshape(X.shape[0], X.shape[1], n_features)
# first input model
visible1 = Input(shape=(n_steps, n_features))
cnn1 = Conv1D(filters=64, kernel_size=2, activation='relu')(visible1)
cnn1 = MaxPooling1D(pool_size=2)(cnn1)
cnn1 = Flatten()(cnn1)
# second input model
visible2 = Input(shape=(n_steps, n_features))
cnn2 = Conv1D(filters=64, kernel_size=2, activation='relu')(visible2)
cnn2 = MaxPooling1D(pool_size=2)(cnn2)
cnn2 = Flatten()(cnn2)
# merge input models
merge = concatenate([cnn1, cnn2])
dense = Dense(50, activation='relu')(merge)
output = Dense(1)(dense)
model = Model(inputs=[visible1, visible2], outputs=output)
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit([X1, X2], y, epochs=1000, verbose=0)
# demonstrate prediction
x_input = array([[80, 85], [90, 95], [100, 105]])
x1 = x_input[:, 0].reshape((1, n_steps, n_features))
x2 = x_input[:, 1].reshape((1, n_steps, n_features))
yhat = model.predict([x1, x2], verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_07_split_parallel_series.py
# multivariate output data prep
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_08_08_cnn_multivariate_parallel_series.py
# multivariate output 1d cnn example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=2, activation='relu', input_shape=(n_steps, n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(50, activation='relu'))
model.add(Dense(n_features))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=3000, verbose=0)
# demonstrate prediction
x_input = array([[70,75,145], [80,85,165], [90,95,185]])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_09_multi_output_cnn_multivariate_parallel_series.py
# multivariate output 1d cnn example
from numpy import array
from numpy import hstack
from keras.models import Model
from keras.layers import Input
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# separate output
y1 = y[:, 0].reshape((y.shape[0], 1))
y2 = y[:, 1].reshape((y.shape[0], 1))
y3 = y[:, 2].reshape((y.shape[0], 1))
# define model
visible = Input(shape=(n_steps, n_features))
cnn = Conv1D(filters=64, kernel_size=2, activation='relu')(visible)
cnn = MaxPooling1D(pool_size=2)(cnn)
cnn = Flatten()(cnn)
cnn = Dense(50, activation='relu')(cnn)
# define output 1
output1 = Dense(1)(cnn)
# define output 2
output2 = Dense(1)(cnn)
# define output 3
output3 = Dense(1)(cnn)
# tie together
model = Model(inputs=visible, outputs=[output1, output2, output3])
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, [y1,y2,y3], epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[70,75,145], [80,85,165], [90,95,185]])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_10_multi_step_dataset.py
# multi-step data preparation
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the sequence
if out_end_ix > len(sequence):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix:out_end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# split into samples
X, y = split_sequence(raw_seq, n_steps_in, n_steps_out)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_08_11_vector_cnn_multi_step.py
# univariate multi-step vector-output 1d cnn example
from numpy import array
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a univariate sequence into samples
def split_sequence(sequence, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the sequence
if out_end_ix > len(sequence):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix:out_end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# split into samples
X, y = split_sequence(raw_seq, n_steps_in, n_steps_out)
# reshape from [samples, timesteps] into [samples, timesteps, features]
n_features = 1
X = X.reshape((X.shape[0], X.shape[1], n_features))
# define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=2, activation='relu', input_shape=(n_steps_in, n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(50, activation='relu'))
model.add(Dense(n_steps_out))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps_in, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_12_multivariate_multistep_dependent_dataset.py
# multivariate multi-step data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out-1
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1:out_end_ix, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_08_13_cnn_multivariate_dependent_multistep.py
# multivariate multi-step 1d cnn example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out-1
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1:out_end_ix, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=2, activation='relu', input_shape=(n_steps_in, n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(50, activation='relu'))
model.add(Dense(n_steps_out))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=2000, verbose=0)
# demonstrate prediction
x_input = array([[70, 75], [80, 85], [90, 95]])
x_input = x_input.reshape((1, n_steps_in, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_08_14_multivariate_multistep_parallel_dataset.py
# multivariate multi-step data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix:out_end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_08_15_cnn_multivariate_parallel_multistep.py
# multivariate output multi-step 1d cnn example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix:out_end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# convert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
# flatten output
n_output = y.shape[1] * y.shape[2]
y = y.reshape((y.shape[0], n_output))
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=2, activation='relu', input_shape=(n_steps_in, n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(50, activation='relu'))
model.add(Dense(n_output))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=7000, verbose=0)
# demonstrate prediction
x_input = array([[60, 65, 125], [70, 75, 145], [80, 85, 165]])
x_input = x_input.reshape((1, n_steps_in, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_01_univariate_dataset.py
# univariate data preparation
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_09_02_vanilla_lstm_univariate.py
# univariate lstm example
from numpy import array
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# reshape from [samples, timesteps] into [samples, timesteps, features]
n_features = 1
X = X.reshape((X.shape[0], X.shape[1], n_features))
# define model
model = Sequential()
model.add(LSTM(50, activation='relu', input_shape=(n_steps, n_features)))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=200, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_03_stacked_lstm_univariate.py
# univariate stacked lstm example
from numpy import array
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
# split a univariate sequence
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# reshape from [samples, timesteps] into [samples, timesteps, features]
n_features = 1
X = X.reshape((X.shape[0], X.shape[1], n_features))
# define model
model = Sequential()
model.add(LSTM(50, activation='relu', return_sequences=True, input_shape=(n_steps, n_features)))
model.add(LSTM(50, activation='relu'))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=200, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_04_bidirectional_lstm_univariate.py
# univariate bidirectional lstm example
from numpy import array
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
from keras.layers import Bidirectional
# split a univariate sequence
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 3
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# reshape from [samples, timesteps] into [samples, timesteps, features]
n_features = 1
X = X.reshape((X.shape[0], X.shape[1], n_features))
# define model
model = Sequential()
model.add(Bidirectional(LSTM(50, activation='relu'), input_shape=(n_steps, n_features)))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=200, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_05_cnn_lstm_univariate.py
# univariate cnn lstm example
from numpy import array
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import TimeDistributed
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 4
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# reshape from [samples, timesteps] into [samples, subsequences, timesteps, features]
n_features = 1
n_seq = 2
n_steps = 2
X = X.reshape((X.shape[0], n_seq, n_steps, n_features))
# define model
model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=1, activation='relu'), input_shape=(None, n_steps, n_features)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(50, activation='relu'))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=500, verbose=0)
# demonstrate prediction
x_input = array([60, 70, 80, 90])
x_input = x_input.reshape((1, n_seq, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_06_convlstm_univariate.py
# univariate convlstm example
from numpy import array
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import ConvLSTM2D
# split a univariate sequence into samples
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps = 4
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# reshape from [samples, timesteps] into [samples, timesteps, rows, columns, features]
n_features = 1
n_seq = 2
n_steps = 2
X = X.reshape((X.shape[0], n_seq, 1, n_steps, n_features))
# define model
model = Sequential()
model.add(ConvLSTM2D(filters=64, kernel_size=(1,2), activation='relu', input_shape=(n_seq, 1, n_steps, n_features)))
model.add(Flatten())
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=500, verbose=0)
# demonstrate prediction
x_input = array([60, 70, 80, 90])
x_input = x_input.reshape((1, n_seq, 1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_07_dependent_series_dataset.py
# multivariate data preparation
from numpy import array
from numpy import hstack
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
print(dataset)
Raw
logicbelief-master-puwork_lstm_chapter_09_08_dependent_series_to_samples.py
# multivariate data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_09_09_vanilla_lstm_multivariate_dependent_series.py
# multivariate lstm example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(LSTM(50, activation='relu', input_shape=(n_steps, n_features)))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=200, verbose=0)
# demonstrate prediction
x_input = array([[80, 85], [90, 95], [100, 105]])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_10_multivariate_parallel_series_dataset.py
# multivariate output data prep
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_09_11_stacked_lstm_multivariate_parallel_series.py
# multivariate output stacked lstm example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps = 3
# convert into input/output
X, y = split_sequences(dataset, n_steps)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(LSTM(100, activation='relu', return_sequences=True, input_shape=(n_steps, n_features)))
model.add(LSTM(100, activation='relu'))
model.add(Dense(n_features))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=400, verbose=0)
# demonstrate prediction
x_input = array([[70,75,145], [80,85,165], [90,95,185]])
x_input = x_input.reshape((1, n_steps, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_12_multi_step_series_dataset.py
# multi-step data preparation
from numpy import array
# split a univariate sequence into samples
def split_sequence(sequence, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the sequence
if out_end_ix > len(sequence):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix:out_end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# split into samples
X, y = split_sequence(raw_seq, n_steps_in, n_steps_out)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_09_13_stacked_lstm_multi_step.py
# univariate multi-step vector-output stacked lstm example
from numpy import array
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
# split a univariate sequence into samples
def split_sequence(sequence, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the sequence
if out_end_ix > len(sequence):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix:out_end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# split into samples
X, y = split_sequence(raw_seq, n_steps_in, n_steps_out)
# reshape from [samples, timesteps] into [samples, timesteps, features]
n_features = 1
X = X.reshape((X.shape[0], X.shape[1], n_features))
# define model
model = Sequential()
model.add(LSTM(100, activation='relu', return_sequences=True, input_shape=(n_steps_in, n_features)))
model.add(LSTM(100, activation='relu'))
model.add(Dense(n_steps_out))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=50, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps_in, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_14_encoder_decoder_lstm_multi_step.py
# univariate multi-step encoder-decoder lstm example
from numpy import array
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
from keras.layers import RepeatVector
from keras.layers import TimeDistributed
# split a univariate sequence into samples
def split_sequence(sequence, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the sequence
if out_end_ix > len(sequence):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix:out_end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
raw_seq = [10, 20, 30, 40, 50, 60, 70, 80, 90]
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# split into samples
X, y = split_sequence(raw_seq, n_steps_in, n_steps_out)
# reshape from [samples, timesteps] into [samples, timesteps, features]
n_features = 1
X = X.reshape((X.shape[0], X.shape[1], n_features))
y = y.reshape((y.shape[0], y.shape[1], n_features))
# define model
model = Sequential()
model.add(LSTM(100, activation='relu', input_shape=(n_steps_in, n_features)))
model.add(RepeatVector(n_steps_out))
model.add(LSTM(100, activation='relu', return_sequences=True))
model.add(TimeDistributed(Dense(1)))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=100, verbose=0)
# demonstrate prediction
x_input = array([70, 80, 90])
x_input = x_input.reshape((1, n_steps_in, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_15_multivariate_dependent_series_multi_step_dataset.py
# multivariate multi-step data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out-1
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1:out_end_ix, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# covert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_09_16_stacked_lstm_dependent_multi_step.py
# multivariate multi-step stacked lstm example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out-1
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :-1], sequences[end_ix-1:out_end_ix, -1]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# covert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(LSTM(100, activation='relu', return_sequences=True, input_shape=(n_steps_in, n_features)))
model.add(LSTM(100, activation='relu'))
model.add(Dense(n_steps_out))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=200, verbose=0)
# demonstrate prediction
x_input = array([[70, 75], [80, 85], [90, 95]])
x_input = x_input.reshape((1, n_steps_in, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_09_17_multivariate_parallel_series_multi_step_dataset.py
# multivariate multi-step data preparation
from numpy import array
from numpy import hstack
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix:out_end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# covert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
print(X.shape, y.shape)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
Raw
logicbelief-master-puwork_lstm_chapter_09_18_encoder_decoder_lstm_parallel_multi_step.py
# multivariate multi-step encoder-decoder lstm example
from numpy import array
from numpy import hstack
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
from keras.layers import RepeatVector
from keras.layers import TimeDistributed
# split a multivariate sequence into samples
def split_sequences(sequences, n_steps_in, n_steps_out):
X, y = list(), list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps_in
out_end_ix = end_ix + n_steps_out
# check if we are beyond the dataset
if out_end_ix > len(sequences):
break
# gather input and output parts of the pattern
seq_x, seq_y = sequences[i:end_ix, :], sequences[end_ix:out_end_ix, :]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# define input sequence
in_seq1 = array([10, 20, 30, 40, 50, 60, 70, 80, 90])
in_seq2 = array([15, 25, 35, 45, 55, 65, 75, 85, 95])
out_seq = array([in_seq1[i]+in_seq2[i] for i in range(len(in_seq1))])
# convert to [rows, columns] structure
in_seq1 = in_seq1.reshape((len(in_seq1), 1))
in_seq2 = in_seq2.reshape((len(in_seq2), 1))
out_seq = out_seq.reshape((len(out_seq), 1))
# horizontally stack columns
dataset = hstack((in_seq1, in_seq2, out_seq))
# choose a number of time steps
n_steps_in, n_steps_out = 3, 2
# covert into input/output
X, y = split_sequences(dataset, n_steps_in, n_steps_out)
# the dataset knows the number of features, e.g. 2
n_features = X.shape[2]
# define model
model = Sequential()
model.add(LSTM(200, activation='relu', input_shape=(n_steps_in, n_features)))
model.add(RepeatVector(n_steps_out))
model.add(LSTM(200, activation='relu', return_sequences=True))
model.add(TimeDistributed(Dense(n_features)))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X, y, epochs=300, verbose=0)
# demonstrate prediction
x_input = array([[60, 65, 125], [70, 75, 145], [80, 85, 165]])
x_input = x_input.reshape((1, n_steps_in, n_features))
yhat = model.predict(x_input, verbose=0)
print(yhat)
Raw
logicbelief-master-puwork_lstm_chapter_11_01_persistence_forecast.py
# example of a one-step naive forecast
def naive_forecast(history, n):
return history[-n]
# define dataset
data = [10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0, 100.0]
print(data)
# test naive forecast
for i in range(1, len(data)+1):
print(naive_forecast(data, i))
Raw
logicbelief-master-puwork_lstm_chapter_11_02_average_forecast.py
# example of an average forecast
from numpy import mean
from numpy import median
# one-step average forecast
def average_forecast(history, config):
n, avg_type = config
# mean of last n values
if avg_type is 'mean':
return mean(history[-n:])
# median of last n values
return median(history[-n:])
# define dataset
data = [10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0, 100.0]
print(data)
# test naive forecast
for i in range(1, len(data)+1):
print(average_forecast(data, (i, 'mean')))
Raw
logicbelief-master-puwork_lstm_chapter_11_03_average_forecast_seasonality.py
# example of an average forecast for seasonal data
from numpy import mean
from numpy import median
# one-step average forecast
def average_forecast(history, config):
n, offset, avg_type = config
values = list()
if offset == 1:
values = history[-n:]
else:
# skip bad configs
if n*offset > len(history):
raise Exception('Config beyond end of data: %d %d' % (n,offset))
# try and collect n values using offset
for i in range(1, n+1):
ix = i * offset
values.append(history[-ix])
# mean of last n values
if avg_type is 'mean':
return mean(values)
# median of last n values
return median(values)
# define dataset
data = [10.0, 20.0, 30.0, 10.0, 20.0, 30.0, 10.0, 20.0, 30.0]
print(data)
# test naive forecast
for i in [1, 2, 3]:
print(average_forecast(data, (i, 3, 'mean')))
Raw
logicbelief-master-puwork_lstm_chapter_11_04_grid_search.py
# grid search simple forecasts
from math import sqrt
from numpy import mean
from numpy import median
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from sklearn.metrics import mean_squared_error
# one-step simple forecast
def simple_forecast(history, config):
n, offset, avg_type = config
# persist value, ignore other config
if avg_type == 'persist':
return history[-n]
# collect values to average
values = list()
if offset == 1:
values = history[-n:]
else:
# skip bad configs
if n*offset > len(history):
raise Exception('Config beyond end of data: %d %d' % (n,offset))
# try and collect n values using offset
for i in range(1, n+1):
ix = i * offset
values.append(history[-ix])
# check if we can average
if len(values) < 2:
raise Exception('Cannot calculate average')
# mean of last n values
if avg_type == 'mean':
return mean(values)
# median of last n values
return median(values)
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = simple_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of simple configs to try
def simple_configs(max_length, offsets=[1]):
configs = list()
for i in range(1, max_length+1):
for o in offsets:
for t in ['persist', 'mean', 'median']:
cfg = [i, o, t]
configs.append(cfg)
return configs
if __name__ == '__main__':
# define dataset
data = [10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0, 100.0]
# data split
n_test = 4
# model configs
max_length = len(data) - n_test
cfg_list = simple_configs(max_length)
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_11_05_load_plot_daily_births.py
# load and plot daily births dataset
from pandas import read_csv
from matplotlib import pyplot
# load
series = read_csv('daily-total-female-births.csv', header=0, index_col=0)
# summarize shape
print(series.shape)
# plot
pyplot.plot(series)
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_11_06_grid_search_daily_births.py
# grid search simple forecast for daily female births
from math import sqrt
from numpy import mean
from numpy import median
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step simple forecast
def simple_forecast(history, config):
n, offset, avg_type = config
# persist value, ignore other config
if avg_type == 'persist':
return history[-n]
# collect values to average
values = list()
if offset == 1:
values = history[-n:]
else:
# skip bad configs
if n*offset > len(history):
raise Exception('Config beyond end of data: %d %d' % (n,offset))
# try and collect n values using offset
for i in range(1, n+1):
ix = i * offset
values.append(history[-ix])
# check if we can average
if len(values) < 2:
raise Exception('Cannot calculate average')
# mean of last n values
if avg_type == 'mean':
return mean(values)
# median of last n values
return median(values)
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = simple_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of simple configs to try
def simple_configs(max_length, offsets=[1]):
configs = list()
for i in range(1, max_length+1):
for o in offsets:
for t in ['persist', 'mean', 'median']:
cfg = [i, o, t]
configs.append(cfg)
return configs
if __name__ == '__main__':
# define dataset
series = read_csv('daily-total-female-births.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 165
# model configs
max_length = len(data) - n_test
cfg_list = simple_configs(max_length)
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_11_07_load_plot_monthly_shampoo.py
# load and plot monthly shampoo sales dataset
from pandas import read_csv
from matplotlib import pyplot
# load
series = read_csv('monthly-shampoo-sales.csv', header=0, index_col=0)
# summarize shape
print(series.shape)
# plot
pyplot.plot(series)
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_11_08_grid_search_shampoo_sales.py
# grid search simple forecast for monthly shampoo sales
from math import sqrt
from numpy import mean
from numpy import median
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step simple forecast
def simple_forecast(history, config):
n, offset, avg_type = config
# persist value, ignore other config
if avg_type == 'persist':
return history[-n]
# collect values to average
values = list()
if offset == 1:
values = history[-n:]
else:
# skip bad configs
if n*offset > len(history):
raise Exception('Config beyond end of data: %d %d' % (n,offset))
# try and collect n values using offset
for i in range(1, n+1):
ix = i * offset
values.append(history[-ix])
# check if we can average
if len(values) < 2:
raise Exception('Cannot calculate average')
# mean of last n values
if avg_type == 'mean':
return mean(values)
# median of last n values
return median(values)
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = simple_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of simple configs to try
def simple_configs(max_length, offsets=[1]):
configs = list()
for i in range(1, max_length+1):
for o in offsets:
for t in ['persist', 'mean', 'median']:
cfg = [i, o, t]
configs.append(cfg)
return configs
if __name__ == '__main__':
# load dataset
series = read_csv('monthly-shampoo-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
max_length = len(data) - n_test
cfg_list = simple_configs(max_length)
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_11_09_load_plot_monthly_mean_temp.py
# load and plot monthly mean temp dataset
from pandas import read_csv
from matplotlib import pyplot
# load
series = read_csv('monthly-mean-temp.csv', header=0, index_col=0)
# summarize shape
print(series.shape)
# plot
pyplot.plot(series)
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_11_10_grid_search_mean_temp.py
# grid search simple forecast for monthly mean temperature
from math import sqrt
from numpy import mean
from numpy import median
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step simple forecast
def simple_forecast(history, config):
n, offset, avg_type = config
# persist value, ignore other config
if avg_type == 'persist':
return history[-n]
# collect values to average
values = list()
if offset == 1:
values = history[-n:]
else:
# skip bad configs
if n*offset > len(history):
raise Exception('Config beyond end of data: %d %d' % (n,offset))
# try and collect n values using offset
for i in range(1, n+1):
ix = i * offset
values.append(history[-ix])
# check if we can average
if len(values) < 2:
raise Exception('Cannot calculate average')
# mean of last n values
if avg_type == 'mean':
return mean(values)
# median of last n values
return median(values)
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = simple_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of simple configs to try
def simple_configs(max_length, offsets=[1]):
configs = list()
for i in range(1, max_length+1):
for o in offsets:
for t in ['persist', 'mean', 'median']:
cfg = [i, o, t]
configs.append(cfg)
return configs
if __name__ == '__main__':
# define dataset
series = read_csv('monthly-mean-temp.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
max_length = len(data) - n_test
cfg_list = simple_configs(max_length, offsets=[1,12])
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_11_11_load_plot_monthly_car_sales.py
# load and plot monthly car sales dataset
from pandas import read_csv
from matplotlib import pyplot
# load
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
# summarize shape
print(series.shape)
# plot
pyplot.plot(series)
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_11_12_grid_search_car_sales.py
# grid search simple forecast for monthly car sales
from math import sqrt
from numpy import mean
from numpy import median
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step simple forecast
def simple_forecast(history, config):
n, offset, avg_type = config
# persist value, ignore other config
if avg_type == 'persist':
return history[-n]
# collect values to average
values = list()
if offset == 1:
values = history[-n:]
else:
# skip bad configs
if n*offset > len(history):
raise Exception('Config beyond end of data: %d %d' % (n,offset))
# try and collect n values using offset
for i in range(1, n+1):
ix = i * offset
values.append(history[-ix])
# check if we can average
if len(values) < 2:
raise Exception('Cannot calculate average')
# mean of last n values
if avg_type == 'mean':
return mean(values)
# median of last n values
return median(values)
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = simple_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of simple configs to try
def simple_configs(max_length, offsets=[1]):
configs = list()
for i in range(1, max_length+1):
for o in offsets:
for t in ['persist', 'mean', 'median']:
cfg = [i, o, t]
configs.append(cfg)
return configs
if __name__ == '__main__':
# define dataset
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
max_length = len(data) - n_test
cfg_list = simple_configs(max_length, offsets=[1,12])
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_11_daily-total-female-births.csv
Date Births
1959-01-01 35
1959-01-02 32
1959-01-03 30
1959-01-04 31
1959-01-05 44
1959-01-06 29
1959-01-07 45
1959-01-08 43
1959-01-09 38
1959-01-10 27
1959-01-11 38
1959-01-12 33
1959-01-13 55
1959-01-14 47
1959-01-15 45
1959-01-16 37
1959-01-17 50
1959-01-18 43
1959-01-19 41
1959-01-20 52
1959-01-21 34
1959-01-22 53
1959-01-23 39
1959-01-24 32
1959-01-25 37
1959-01-26 43
1959-01-27 39
1959-01-28 35
1959-01-29 44
1959-01-30 38
1959-01-31 24
1959-02-01 23
1959-02-02 31
1959-02-03 44
1959-02-04 38
1959-02-05 50
1959-02-06 38
1959-02-07 51
1959-02-08 31
1959-02-09 31
1959-02-10 51
1959-02-11 36
1959-02-12 45
1959-02-13 51
1959-02-14 34
1959-02-15 52
1959-02-16 47
1959-02-17 45
1959-02-18 46
1959-02-19 39
1959-02-20 48
1959-02-21 37
1959-02-22 35
1959-02-23 52
1959-02-24 42
1959-02-25 45
1959-02-26 39
1959-02-27 37
1959-02-28 30
1959-03-01 35
1959-03-02 28
1959-03-03 45
1959-03-04 34
1959-03-05 36
1959-03-06 50
1959-03-07 44
1959-03-08 39
1959-03-09 32
1959-03-10 39
1959-03-11 45
1959-03-12 43
1959-03-13 39
1959-03-14 31
1959-03-15 27
1959-03-16 30
1959-03-17 42
1959-03-18 46
1959-03-19 41
1959-03-20 36
1959-03-21 45
1959-03-22 46
1959-03-23 43
1959-03-24 38
1959-03-25 34
1959-03-26 35
1959-03-27 56
1959-03-28 36
1959-03-29 32
1959-03-30 50
1959-03-31 41
1959-04-01 39
1959-04-02 41
1959-04-03 47
1959-04-04 34
1959-04-05 36
1959-04-06 33
1959-04-07 35
1959-04-08 38
1959-04-09 38
1959-04-10 34
1959-04-11 53
1959-04-12 34
1959-04-13 34
1959-04-14 38
1959-04-15 35
1959-04-16 32
1959-04-17 42
1959-04-18 34
1959-04-19 46
1959-04-20 30
1959-04-21 46
1959-04-22 45
1959-04-23 54
1959-04-24 34
1959-04-25 37
1959-04-26 35
1959-04-27 40
1959-04-28 42
1959-04-29 58
1959-04-30 51
1959-05-01 32
1959-05-02 35
1959-05-03 38
1959-05-04 33
1959-05-05 39
1959-05-06 47
1959-05-07 38
1959-05-08 52
1959-05-09 30
1959-05-10 34
1959-05-11 40
1959-05-12 35
1959-05-13 42
1959-05-14 41
1959-05-15 42
1959-05-16 38
1959-05-17 24
1959-05-18 34
1959-05-19 43
1959-05-20 36
1959-05-21 55
1959-05-22 41
1959-05-23 45
1959-05-24 41
1959-05-25 37
1959-05-26 43
1959-05-27 39
1959-05-28 33
1959-05-29 43
1959-05-30 40
1959-05-31 38
1959-06-01 45
1959-06-02 46
1959-06-03 34
1959-06-04 35
1959-06-05 48
1959-06-06 51
1959-06-07 36
1959-06-08 33
1959-06-09 46
1959-06-10 42
1959-06-11 48
1959-06-12 34
1959-06-13 41
1959-06-14 35
1959-06-15 40
1959-06-16 34
1959-06-17 30
1959-06-18 36
1959-06-19 40
1959-06-20 39
1959-06-21 45
1959-06-22 38
1959-06-23 47
1959-06-24 33
1959-06-25 30
1959-06-26 42
1959-06-27 43
1959-06-28 41
1959-06-29 41
1959-06-30 59
1959-07-01 43
1959-07-02 45
1959-07-03 38
1959-07-04 37
1959-07-05 45
1959-07-06 42
1959-07-07 57
1959-07-08 46
1959-07-09 51
1959-07-10 41
1959-07-11 47
1959-07-12 26
1959-07-13 35
1959-07-14 44
1959-07-15 41
1959-07-16 42
1959-07-17 36
1959-07-18 45
1959-07-19 45
1959-07-20 45
1959-07-21 47
1959-07-22 38
1959-07-23 42
1959-07-24 35
1959-07-25 36
1959-07-26 39
1959-07-27 45
1959-07-28 43
1959-07-29 47
1959-07-30 36
1959-07-31 41
1959-08-01 50
1959-08-02 39
1959-08-03 41
1959-08-04 46
1959-08-05 64
1959-08-06 45
1959-08-07 34
1959-08-08 38
1959-08-09 44
1959-08-10 48
1959-08-11 46
1959-08-12 44
1959-08-13 37
1959-08-14 39
1959-08-15 44
1959-08-16 45
1959-08-17 33
1959-08-18 44
1959-08-19 38
1959-08-20 46
1959-08-21 46
1959-08-22 40
1959-08-23 39
1959-08-24 44
1959-08-25 48
1959-08-26 50
1959-08-27 41
1959-08-28 42
1959-08-29 51
1959-08-30 41
1959-08-31 44
1959-09-01 38
1959-09-02 68
1959-09-03 40
1959-09-04 42
1959-09-05 51
1959-09-06 44
1959-09-07 45
1959-09-08 36
1959-09-09 57
1959-09-10 44
1959-09-11 42
1959-09-12 53
1959-09-13 42
1959-09-14 34
1959-09-15 40
1959-09-16 56
1959-09-17 44
1959-09-18 53
1959-09-19 55
1959-09-20 39
1959-09-21 59
1959-09-22 55
1959-09-23 73
1959-09-24 55
1959-09-25 44
1959-09-26 43
1959-09-27 40
1959-09-28 47
1959-09-29 51
1959-09-30 56
1959-10-01 49
1959-10-02 54
1959-10-03 56
1959-10-04 47
1959-10-05 44
1959-10-06 43
1959-10-07 42
1959-10-08 45
1959-10-09 50
1959-10-10 48
1959-10-11 43
1959-10-12 40
1959-10-13 59
1959-10-14 41
1959-10-15 42
1959-10-16 51
1959-10-17 49
1959-10-18 45
1959-10-19 43
1959-10-20 42
1959-10-21 38
1959-10-22 47
1959-10-23 38
1959-10-24 36
1959-10-25 42
1959-10-26 35
1959-10-27 28
1959-10-28 44
1959-10-29 36
1959-10-30 45
1959-10-31 46
1959-11-01 48
1959-11-02 49
1959-11-03 43
1959-11-04 42
1959-11-05 59
1959-11-06 45
1959-11-07 52
1959-11-08 46
1959-11-09 42
1959-11-10 40
1959-11-11 40
1959-11-12 45
1959-11-13 35
1959-11-14 35
1959-11-15 40
1959-11-16 39
1959-11-17 33
1959-11-18 42
1959-11-19 47
1959-11-20 51
1959-11-21 44
1959-11-22 40
1959-11-23 57
1959-11-24 49
1959-11-25 45
1959-11-26 49
1959-11-27 51
1959-11-28 46
1959-11-29 44
1959-11-30 52
1959-12-01 45
1959-12-02 32
1959-12-03 46
1959-12-04 41
1959-12-05 34
1959-12-06 33
1959-12-07 36
1959-12-08 49
1959-12-09 43
1959-12-10 43
1959-12-11 34
1959-12-12 39
1959-12-13 35
1959-12-14 52
1959-12-15 47
1959-12-16 52
1959-12-17 39
1959-12-18 40
1959-12-19 42
1959-12-20 42
1959-12-21 53
1959-12-22 39
1959-12-23 40
1959-12-24 38
1959-12-25 44
1959-12-26 34
1959-12-27 37
1959-12-28 52
1959-12-29 48
1959-12-30 55
1959-12-31 50
Raw
logicbelief-master-puwork_lstm_chapter_11_monthly-car-sales.csv
Month Car Sales
1960-01 6550
1960-02 8728
1960-03 12026
1960-04 14395
1960-05 14587
1960-06 13791
1960-07 9498
1960-08 8251
1960-09 7049
1960-10 9545
1960-11 9364
1960-12 8456
1961-01 7237
1961-02 9374
1961-03 11837
1961-04 13784
1961-05 15926
1961-06 13821
1961-07 11143
1961-08 7975
1961-09 7610
1961-10 10015
1961-11 12759
1961-12 8816
1962-01 10677
1962-02 10947
1962-03 15200
1962-04 17010
1962-05 20900
1962-06 16205
1962-07 12143
1962-08 8997
1962-09 5568
1962-10 11474
1962-11 12256
1962-12 10583
1963-01 10862
1963-02 10965
1963-03 14405
1963-04 20379
1963-05 20128
1963-06 17816
1963-07 12268
1963-08 8642
1963-09 7962
1963-10 13932
1963-11 15936
1963-12 12628
1964-01 12267
1964-02 12470
1964-03 18944
1964-04 21259
1964-05 22015
1964-06 18581
1964-07 15175
1964-08 10306
1964-09 10792
1964-10 14752
1964-11 13754
1964-12 11738
1965-01 12181
1965-02 12965
1965-03 19990
1965-04 23125
1965-05 23541
1965-06 21247
1965-07 15189
1965-08 14767
1965-09 10895
1965-10 17130
1965-11 17697
1965-12 16611
1966-01 12674
1966-02 12760
1966-03 20249
1966-04 22135
1966-05 20677
1966-06 19933
1966-07 15388
1966-08 15113
1966-09 13401
1966-10 16135
1966-11 17562
1966-12 14720
1967-01 12225
1967-02 11608
1967-03 20985
1967-04 19692
1967-05 24081
1967-06 22114
1967-07 14220
1967-08 13434
1967-09 13598
1967-10 17187
1967-11 16119
1967-12 13713
1968-01 13210
1968-02 14251
1968-03 20139
1968-04 21725
1968-05 26099
1968-06 21084
1968-07 18024
1968-08 16722
1968-09 14385
1968-10 21342
1968-11 17180
1968-12 14577
Raw
logicbelief-master-puwork_lstm_chapter_11_monthly-mean-temp.csv
Month Temperature
1920-01 40.6
1920-02 40.8
1920-03 44.4
1920-04 46.7
1920-05 54.1
1920-06 58.5
1920-07 57.7
1920-08 56.4
1920-09 54.3
1920-10 50.5
1920-11 42.9
1920-12 39.8
1921-01 44.2
1921-02 39.8
1921-03 45.1
1921-04 47.0
1921-05 54.1
1921-06 58.7
1921-07 66.3
1921-08 59.9
1921-09 57.0
1921-10 54.2
1921-11 39.7
1921-12 42.8
1922-01 37.5
1922-02 38.7
1922-03 39.5
1922-04 42.1
1922-05 55.7
1922-06 57.8
1922-07 56.8
1922-08 54.3
1922-09 54.3
1922-10 47.1
1922-11 41.8
1922-12 41.7
1923-01 41.8
1923-02 40.1
1923-03 42.9
1923-04 45.8
1923-05 49.2
1923-06 52.7
1923-07 64.2
1923-08 59.6
1923-09 54.4
1923-10 49.2
1923-11 36.6
1923-12 37.6
1924-01 39.3
1924-02 37.5
1924-03 38.3
1924-04 45.5
1924-05 53.2
1924-06 57.7
1924-07 60.8
1924-08 58.2
1924-09 56.4
1924-10 49.8
1924-11 44.4
1924-12 43.6
1925-01 40.0
1925-02 40.5
1925-03 40.8
1925-04 45.1
1925-05 53.8
1925-06 59.4
1925-07 63.5
1925-08 61.0
1925-09 53.0
1925-10 50.0
1925-11 38.1
1925-12 36.3
1926-01 39.2
1926-02 43.4
1926-03 43.4
1926-04 48.9
1926-05 50.6
1926-06 56.8
1926-07 62.5
1926-08 62.0
1926-09 57.5
1926-10 46.7
1926-11 41.6
1926-12 39.8
1927-01 39.4
1927-02 38.5
1927-03 45.3
1927-04 47.1
1927-05 51.7
1927-06 55.0
1927-07 60.4
1927-08 60.5
1927-09 54.7
1927-10 50.3
1927-11 42.3
1927-12 35.2
1928-01 40.8
1928-02 41.1
1928-03 42.8
1928-04 47.3
1928-05 50.9
1928-06 56.4
1928-07 62.2
1928-08 60.5
1928-09 55.4
1928-10 50.2
1928-11 43.0
1928-12 37.3
1929-01 34.8
1929-02 31.3
1929-03 41.0
1929-04 43.9
1929-05 53.1
1929-06 56.9
1929-07 62.5
1929-08 60.3
1929-09 59.8
1929-10 49.2
1929-11 42.9
1929-12 41.9
1930-01 41.6
1930-02 37.1
1930-03 41.2
1930-04 46.9
1930-05 51.2
1930-06 60.4
1930-07 60.1
1930-08 61.6
1930-09 57.0
1930-10 50.9
1930-11 43.0
1930-12 38.8
1931-01 37.1
1931-02 38.4
1931-03 38.4
1931-04 46.5
1931-05 53.5
1931-06 58.4
1931-07 60.6
1931-08 58.2
1931-09 53.8
1931-10 46.6
1931-11 45.5
1931-12 40.6
1932-01 42.4
1932-02 38.4
1932-03 40.3
1932-04 44.6
1932-05 50.9
1932-06 57.0
1932-07 62.1
1932-08 63.5
1932-09 56.2
1932-10 47.3
1932-11 43.6
1932-12 41.8
1933-01 36.2
1933-02 39.3
1933-03 44.5
1933-04 48.7
1933-05 54.2
1933-06 60.8
1933-07 65.5
1933-08 64.9
1933-09 60.1
1933-10 50.2
1933-11 42.1
1933-12 35.6
1934-01 39.4
1934-02 38.2
1934-03 40.4
1934-04 46.9
1934-05 53.4
1934-06 59.6
1934-07 66.5
1934-08 60.4
1934-09 59.2
1934-10 51.2
1934-11 42.8
1934-12 45.8
1935-01 40.4
1935-02 42.6
1935-03 43.5
1935-04 47.1
1935-05 50.0
1935-06 60.5
1935-07 64.6
1935-08 64.0
1935-09 56.8
1935-10 48.6
1935-11 44.2
1935-12 36.4
1936-01 37.3
1936-02 35.0
1936-03 44.0
1936-04 43.9
1936-05 52.7
1936-06 58.6
1936-07 60.0
1936-08 61.1
1936-09 58.1
1936-10 49.6
1936-11 41.6
1936-12 41.3
1937-01 40.8
1937-02 41.0
1937-03 38.4
1937-04 47.4
1937-05 54.1
1937-06 58.6
1937-07 61.4
1937-08 61.8
1937-09 56.3
1937-10 50.9
1937-11 41.4
1937-12 37.1
1938-01 42.1
1938-02 41.2
1938-03 47.3
1938-04 46.6
1938-05 52.4
1938-06 59.0
1938-07 59.6
1938-08 60.4
1938-09 57.0
1938-10 50.7
1938-11 47.8
1938-12 39.2
1939-01 39.4
1939-02 40.9
1939-03 42.4
1939-04 47.8
1939-05 52.4
1939-06 58.0
1939-07 60.7
1939-08 61.8
1939-09 58.2
1939-10 46.7
1939-11 46.6
1939-12 37.8
Raw
logicbelief-master-puwork_lstm_chapter_11_monthly-shampoo-sales.csv
Month Sales
1-01 266.0
1-02 145.9
1-03 183.1
1-04 119.3
1-05 180.3
1-06 168.5
1-07 231.8
1-08 224.5
1-09 192.8
1-10 122.9
1-11 336.5
1-12 185.9
2-01 194.3
2-02 149.5
2-03 210.1
2-04 273.3
2-05 191.4
2-06 287.0
2-07 226.0
2-08 303.6
2-09 289.9
2-10 421.6
2-11 264.5
2-12 342.3
3-01 339.7
3-02 440.4
3-03 315.9
3-04 439.3
3-05 401.3
3-06 437.4
3-07 575.5
3-08 407.6
3-09 682.0
3-10 475.3
3-11 581.3
3-12 646.9
Raw
logicbelief-master-puwork_lstm_chapter_12_01_grid_search.py
# grid search holt winter's exponential smoothing
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from sklearn.metrics import mean_squared_error
from numpy import array
# one-step Holt Winter’s Exponential Smoothing forecast
def exp_smoothing_forecast(history, config):
t,d,s,p,b,r = config
# define model
history = array(history)
model = ExponentialSmoothing(history, trend=t, damped=d, seasonal=s, seasonal_periods=p)
# fit model
model_fit = model.fit(optimized=True, use_boxcox=b, remove_bias=r)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = exp_smoothing_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of exponential smoothing configs to try
def exp_smoothing_configs(seasonal=[None]):
models = list()
# define config lists
t_params = ['add', 'mul', None]
d_params = [True, False]
s_params = ['add', 'mul', None]
p_params = seasonal
b_params = [True, False]
r_params = [True, False]
# create config instances
for t in t_params:
for d in d_params:
for s in s_params:
for p in p_params:
for b in b_params:
for r in r_params:
cfg = [t,d,s,p,b,r]
models.append(cfg)
return models
if __name__ == '__main__':
# define dataset
data = [10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0, 100.0]
print(data)
# data split
n_test = 4
# model configs
cfg_list = exp_smoothing_configs()
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_12_02_grid_search_daily_births.py
# grid search ets models for daily female births
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from sklearn.metrics import mean_squared_error
from pandas import read_csv
from numpy import array
# one-step Holt Winter’s Exponential Smoothing forecast
def exp_smoothing_forecast(history, config):
t,d,s,p,b,r = config
# define model
history = array(history)
model = ExponentialSmoothing(history, trend=t, damped=d, seasonal=s, seasonal_periods=p)
# fit model
model_fit = model.fit(optimized=True, use_boxcox=b, remove_bias=r)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = exp_smoothing_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of exponential smoothing configs to try
def exp_smoothing_configs(seasonal=[None]):
models = list()
# define config lists
t_params = ['add', 'mul', None]
d_params = [True, False]
s_params = ['add', 'mul', None]
p_params = seasonal
b_params = [True, False]
r_params = [True, False]
# create config instances
for t in t_params:
for d in d_params:
for s in s_params:
for p in p_params:
for b in b_params:
for r in r_params:
cfg = [t,d,s,p,b,r]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('daily-total-female-births.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 165
# model configs
cfg_list = exp_smoothing_configs()
# grid search
scores = grid_search(data[:,0], cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_12_03_grid_search_monthly_shampoo_sales.py
# grid search ets models for monthly shampoo sales
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from sklearn.metrics import mean_squared_error
from pandas import read_csv
from numpy import array
# one-step Holt Winter’s Exponential Smoothing forecast
def exp_smoothing_forecast(history, config):
t,d,s,p,b,r = config
# define model
history = array(history)
model = ExponentialSmoothing(history, trend=t, damped=d, seasonal=s, seasonal_periods=p)
# fit model
model_fit = model.fit(optimized=True, use_boxcox=b, remove_bias=r)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = exp_smoothing_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of exponential smoothing configs to try
def exp_smoothing_configs(seasonal=[None]):
models = list()
# define config lists
t_params = ['add', 'mul', None]
d_params = [True, False]
s_params = ['add', 'mul', None]
p_params = seasonal
b_params = [True, False]
r_params = [True, False]
# create config instances
for t in t_params:
for d in d_params:
for s in s_params:
for p in p_params:
for b in b_params:
for r in r_params:
cfg = [t,d,s,p,b,r]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('monthly-shampoo-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = exp_smoothing_configs()
# grid search
scores = grid_search(data[:,0], cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_12_04_grid_search_monthly_mean_temp.py
# grid search ets hyperparameters for monthly mean temp dataset
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from sklearn.metrics import mean_squared_error
from pandas import read_csv
from numpy import array
# one-step Holt Winter’s Exponential Smoothing forecast
def exp_smoothing_forecast(history, config):
t,d,s,p,b,r = config
# define model
history = array(history)
model = ExponentialSmoothing(history, trend=t, damped=d, seasonal=s, seasonal_periods=p)
# fit model
model_fit = model.fit(optimized=True, use_boxcox=b, remove_bias=r)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = exp_smoothing_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of exponential smoothing configs to try
def exp_smoothing_configs(seasonal=[None]):
models = list()
# define config lists
t_params = ['add', 'mul', None]
d_params = [True, False]
s_params = ['add', 'mul', None]
p_params = seasonal
b_params = [True, False]
r_params = [True, False]
# create config instances
for t in t_params:
for d in d_params:
for s in s_params:
for p in p_params:
for b in b_params:
for r in r_params:
cfg = [t,d,s,p,b,r]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('monthly-mean-temp.csv', header=0, index_col=0)
data = series.values
# trim dataset to 5 years
data = data[-(5*12):]
# data split
n_test = 12
# model configs
cfg_list = exp_smoothing_configs(seasonal=[0,12])
# grid search
scores = grid_search(data[:,0], cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_12_05_grid_search_monthly_car_sales.py
# grid search ets models for monthly car sales
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from sklearn.metrics import mean_squared_error
from pandas import read_csv
from numpy import array
# one-step Holt Winter’s Exponential Smoothing forecast
def exp_smoothing_forecast(history, config):
t,d,s,p,b,r = config
# define model
history = array(history)
model = ExponentialSmoothing(history, trend=t, damped=d, seasonal=s, seasonal_periods=p)
# fit model
model_fit = model.fit(optimized=True, use_boxcox=b, remove_bias=r)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = exp_smoothing_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of exponential smoothing configs to try
def exp_smoothing_configs(seasonal=[None]):
models = list()
# define config lists
t_params = ['add', 'mul', None]
d_params = [True, False]
s_params = ['add', 'mul', None]
p_params = seasonal
b_params = [True, False]
r_params = [True, False]
# create config instances
for t in t_params:
for d in d_params:
for s in s_params:
for p in p_params:
for b in b_params:
for r in r_params:
cfg = [t,d,s,p,b,r]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = exp_smoothing_configs(seasonal=[0,6,12])
# grid search
scores = grid_search(data[:,0], cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_12_daily-total-female-births.csv
1959-01-01 35
1959-01-02 32
1959-01-03 30
1959-01-04 31
1959-01-05 44
1959-01-06 29
1959-01-07 45
1959-01-08 43
1959-01-09 38
1959-01-10 27
1959-01-11 38
1959-01-12 33
1959-01-13 55
1959-01-14 47
1959-01-15 45
1959-01-16 37
1959-01-17 50
1959-01-18 43
1959-01-19 41
1959-01-20 52
1959-01-21 34
1959-01-22 53
1959-01-23 39
1959-01-24 32
1959-01-25 37
1959-01-26 43
1959-01-27 39
1959-01-28 35
1959-01-29 44
1959-01-30 38
1959-01-31 24
1959-02-01 23
1959-02-02 31
1959-02-03 44
1959-02-04 38
1959-02-05 50
1959-02-06 38
1959-02-07 51
1959-02-08 31
1959-02-09 31
1959-02-10 51
1959-02-11 36
1959-02-12 45
1959-02-13 51
1959-02-14 34
1959-02-15 52
1959-02-16 47
1959-02-17 45
1959-02-18 46
1959-02-19 39
1959-02-20 48
1959-02-21 37
1959-02-22 35
1959-02-23 52
1959-02-24 42
1959-02-25 45
1959-02-26 39
1959-02-27 37
1959-02-28 30
1959-03-01 35
1959-03-02 28
1959-03-03 45
1959-03-04 34
1959-03-05 36
1959-03-06 50
1959-03-07 44
1959-03-08 39
1959-03-09 32
1959-03-10 39
1959-03-11 45
1959-03-12 43
1959-03-13 39
1959-03-14 31
1959-03-15 27
1959-03-16 30
1959-03-17 42
1959-03-18 46
1959-03-19 41
1959-03-20 36
1959-03-21 45
1959-03-22 46
1959-03-23 43
1959-03-24 38
1959-03-25 34
1959-03-26 35
1959-03-27 56
1959-03-28 36
1959-03-29 32
1959-03-30 50
1959-03-31 41
1959-04-01 39
1959-04-02 41
1959-04-03 47
1959-04-04 34
1959-04-05 36
1959-04-06 33
1959-04-07 35
1959-04-08 38
1959-04-09 38
1959-04-10 34
1959-04-11 53
1959-04-12 34
1959-04-13 34
1959-04-14 38
1959-04-15 35
1959-04-16 32
1959-04-17 42
1959-04-18 34
1959-04-19 46
1959-04-20 30
1959-04-21 46
1959-04-22 45
1959-04-23 54
1959-04-24 34
1959-04-25 37
1959-04-26 35
1959-04-27 40
1959-04-28 42
1959-04-29 58
1959-04-30 51
1959-05-01 32
1959-05-02 35
1959-05-03 38
1959-05-04 33
1959-05-05 39
1959-05-06 47
1959-05-07 38
1959-05-08 52
1959-05-09 30
1959-05-10 34
1959-05-11 40
1959-05-12 35
1959-05-13 42
1959-05-14 41
1959-05-15 42
1959-05-16 38
1959-05-17 24
1959-05-18 34
1959-05-19 43
1959-05-20 36
1959-05-21 55
1959-05-22 41
1959-05-23 45
1959-05-24 41
1959-05-25 37
1959-05-26 43
1959-05-27 39
1959-05-28 33
1959-05-29 43
1959-05-30 40
1959-05-31 38
1959-06-01 45
1959-06-02 46
1959-06-03 34
1959-06-04 35
1959-06-05 48
1959-06-06 51
1959-06-07 36
1959-06-08 33
1959-06-09 46
1959-06-10 42
1959-06-11 48
1959-06-12 34
1959-06-13 41
1959-06-14 35
1959-06-15 40
1959-06-16 34
1959-06-17 30
1959-06-18 36
1959-06-19 40
1959-06-20 39
1959-06-21 45
1959-06-22 38
1959-06-23 47
1959-06-24 33
1959-06-25 30
1959-06-26 42
1959-06-27 43
1959-06-28 41
1959-06-29 41
1959-06-30 59
1959-07-01 43
1959-07-02 45
1959-07-03 38
1959-07-04 37
1959-07-05 45
1959-07-06 42
1959-07-07 57
1959-07-08 46
1959-07-09 51
1959-07-10 41
1959-07-11 47
1959-07-12 26
1959-07-13 35
1959-07-14 44
1959-07-15 41
1959-07-16 42
1959-07-17 36
1959-07-18 45
1959-07-19 45
1959-07-20 45
1959-07-21 47
1959-07-22 38
1959-07-23 42
1959-07-24 35
1959-07-25 36
1959-07-26 39
1959-07-27 45
1959-07-28 43
1959-07-29 47
1959-07-30 36
1959-07-31 41
1959-08-01 50
1959-08-02 39
1959-08-03 41
1959-08-04 46
1959-08-05 64
1959-08-06 45
1959-08-07 34
1959-08-08 38
1959-08-09 44
1959-08-10 48
1959-08-11 46
1959-08-12 44
1959-08-13 37
1959-08-14 39
1959-08-15 44
1959-08-16 45
1959-08-17 33
1959-08-18 44
1959-08-19 38
1959-08-20 46
1959-08-21 46
1959-08-22 40
1959-08-23 39
1959-08-24 44
1959-08-25 48
1959-08-26 50
1959-08-27 41
1959-08-28 42
1959-08-29 51
1959-08-30 41
1959-08-31 44
1959-09-01 38
1959-09-02 68
1959-09-03 40
1959-09-04 42
1959-09-05 51
1959-09-06 44
1959-09-07 45
1959-09-08 36
1959-09-09 57
1959-09-10 44
1959-09-11 42
1959-09-12 53
1959-09-13 42
1959-09-14 34
1959-09-15 40
1959-09-16 56
1959-09-17 44
1959-09-18 53
1959-09-19 55
1959-09-20 39
1959-09-21 59
1959-09-22 55
1959-09-23 73
1959-09-24 55
1959-09-25 44
1959-09-26 43
1959-09-27 40
1959-09-28 47
1959-09-29 51
1959-09-30 56
1959-10-01 49
1959-10-02 54
1959-10-03 56
1959-10-04 47
1959-10-05 44
1959-10-06 43
1959-10-07 42
1959-10-08 45
1959-10-09 50
1959-10-10 48
1959-10-11 43
1959-10-12 40
1959-10-13 59
1959-10-14 41
1959-10-15 42
1959-10-16 51
1959-10-17 49
1959-10-18 45
1959-10-19 43
1959-10-20 42
1959-10-21 38
1959-10-22 47
1959-10-23 38
1959-10-24 36
1959-10-25 42
1959-10-26 35
1959-10-27 28
1959-10-28 44
1959-10-29 36
1959-10-30 45
1959-10-31 46
1959-11-01 48
1959-11-02 49
1959-11-03 43
1959-11-04 42
1959-11-05 59
1959-11-06 45
1959-11-07 52
1959-11-08 46
1959-11-09 42
1959-11-10 40
1959-11-11 40
1959-11-12 45
1959-11-13 35
1959-11-14 35
1959-11-15 40
1959-11-16 39
1959-11-17 33
1959-11-18 42
1959-11-19 47
1959-11-20 51
1959-11-21 44
1959-11-22 40
1959-11-23 57
1959-11-24 49
1959-11-25 45
1959-11-26 49
1959-11-27 51
1959-11-28 46
1959-11-29 44
1959-11-30 52
1959-12-01 45
1959-12-02 32
1959-12-03 46
1959-12-04 41
1959-12-05 34
1959-12-06 33
1959-12-07 36
1959-12-08 49
1959-12-09 43
1959-12-10 43
1959-12-11 34
1959-12-12 39
1959-12-13 35
1959-12-14 52
1959-12-15 47
1959-12-16 52
1959-12-17 39
1959-12-18 40
1959-12-19 42
1959-12-20 42
1959-12-21 53
1959-12-22 39
1959-12-23 40
1959-12-24 38
1959-12-25 44
1959-12-26 34
1959-12-27 37
1959-12-28 52
1959-12-29 48
1959-12-30 55
1959-12-31 50
Raw
logicbelief-master-puwork_lstm_chapter_12_monthly-car-sales.csv
1960-01 6550
1960-02 8728
1960-03 12026
1960-04 14395
1960-05 14587
1960-06 13791
1960-07 9498
1960-08 8251
1960-09 7049
1960-10 9545
1960-11 9364
1960-12 8456
1961-01 7237
1961-02 9374
1961-03 11837
1961-04 13784
1961-05 15926
1961-06 13821
1961-07 11143
1961-08 7975
1961-09 7610
1961-10 10015
1961-11 12759
1961-12 8816
1962-01 10677
1962-02 10947
1962-03 15200
1962-04 17010
1962-05 20900
1962-06 16205
1962-07 12143
1962-08 8997
1962-09 5568
1962-10 11474
1962-11 12256
1962-12 10583
1963-01 10862
1963-02 10965
1963-03 14405
1963-04 20379
1963-05 20128
1963-06 17816
1963-07 12268
1963-08 8642
1963-09 7962
1963-10 13932
1963-11 15936
1963-12 12628
1964-01 12267
1964-02 12470
1964-03 18944
1964-04 21259
1964-05 22015
1964-06 18581
1964-07 15175
1964-08 10306
1964-09 10792
1964-10 14752
1964-11 13754
1964-12 11738
1965-01 12181
1965-02 12965
1965-03 19990
1965-04 23125
1965-05 23541
1965-06 21247
1965-07 15189
1965-08 14767
1965-09 10895
1965-10 17130
1965-11 17697
1965-12 16611
1966-01 12674
1966-02 12760
1966-03 20249
1966-04 22135
1966-05 20677
1966-06 19933
1966-07 15388
1966-08 15113
1966-09 13401
1966-10 16135
1966-11 17562
1966-12 14720
1967-01 12225
1967-02 11608
1967-03 20985
1967-04 19692
1967-05 24081
1967-06 22114
1967-07 14220
1967-08 13434
1967-09 13598
1967-10 17187
1967-11 16119
1967-12 13713
1968-01 13210
1968-02 14251
1968-03 20139
1968-04 21725
1968-05 26099
1968-06 21084
1968-07 18024
1968-08 16722
1968-09 14385
1968-10 21342
1968-11 17180
1968-12 14577
Raw
logicbelief-master-puwork_lstm_chapter_12_monthly-mean-temp.csv
1920-01 40.6
1920-02 40.8
1920-03 44.4
1920-04 46.7
1920-05 54.1
1920-06 58.5
1920-07 57.7
1920-08 56.4
1920-09 54.3
1920-10 50.5
1920-11 42.9
1920-12 39.8
1921-01 44.2
1921-02 39.8
1921-03 45.1
1921-04 47.0
1921-05 54.1
1921-06 58.7
1921-07 66.3
1921-08 59.9
1921-09 57.0
1921-10 54.2
1921-11 39.7
1921-12 42.8
1922-01 37.5
1922-02 38.7
1922-03 39.5
1922-04 42.1
1922-05 55.7
1922-06 57.8
1922-07 56.8
1922-08 54.3
1922-09 54.3
1922-10 47.1
1922-11 41.8
1922-12 41.7
1923-01 41.8
1923-02 40.1
1923-03 42.9
1923-04 45.8
1923-05 49.2
1923-06 52.7
1923-07 64.2
1923-08 59.6
1923-09 54.4
1923-10 49.2
1923-11 36.6
1923-12 37.6
1924-01 39.3
1924-02 37.5
1924-03 38.3
1924-04 45.5
1924-05 53.2
1924-06 57.7
1924-07 60.8
1924-08 58.2
1924-09 56.4
1924-10 49.8
1924-11 44.4
1924-12 43.6
1925-01 40.0
1925-02 40.5
1925-03 40.8
1925-04 45.1
1925-05 53.8
1925-06 59.4
1925-07 63.5
1925-08 61.0
1925-09 53.0
1925-10 50.0
1925-11 38.1
1925-12 36.3
1926-01 39.2
1926-02 43.4
1926-03 43.4
1926-04 48.9
1926-05 50.6
1926-06 56.8
1926-07 62.5
1926-08 62.0
1926-09 57.5
1926-10 46.7
1926-11 41.6
1926-12 39.8
1927-01 39.4
1927-02 38.5
1927-03 45.3
1927-04 47.1
1927-05 51.7
1927-06 55.0
1927-07 60.4
1927-08 60.5
1927-09 54.7
1927-10 50.3
1927-11 42.3
1927-12 35.2
1928-01 40.8
1928-02 41.1
1928-03 42.8
1928-04 47.3
1928-05 50.9
1928-06 56.4
1928-07 62.2
1928-08 60.5
1928-09 55.4
1928-10 50.2
1928-11 43.0
1928-12 37.3
1929-01 34.8
1929-02 31.3
1929-03 41.0
1929-04 43.9
1929-05 53.1
1929-06 56.9
1929-07 62.5
1929-08 60.3
1929-09 59.8
1929-10 49.2
1929-11 42.9
1929-12 41.9
1930-01 41.6
1930-02 37.1
1930-03 41.2
1930-04 46.9
1930-05 51.2
1930-06 60.4
1930-07 60.1
1930-08 61.6
1930-09 57.0
1930-10 50.9
1930-11 43.0
1930-12 38.8
1931-01 37.1
1931-02 38.4
1931-03 38.4
1931-04 46.5
1931-05 53.5
1931-06 58.4
1931-07 60.6
1931-08 58.2
1931-09 53.8
1931-10 46.6
1931-11 45.5
1931-12 40.6
1932-01 42.4
1932-02 38.4
1932-03 40.3
1932-04 44.6
1932-05 50.9
1932-06 57.0
1932-07 62.1
1932-08 63.5
1932-09 56.2
1932-10 47.3
1932-11 43.6
1932-12 41.8
1933-01 36.2
1933-02 39.3
1933-03 44.5
1933-04 48.7
1933-05 54.2
1933-06 60.8
1933-07 65.5
1933-08 64.9
1933-09 60.1
1933-10 50.2
1933-11 42.1
1933-12 35.6
1934-01 39.4
1934-02 38.2
1934-03 40.4
1934-04 46.9
1934-05 53.4
1934-06 59.6
1934-07 66.5
1934-08 60.4
1934-09 59.2
1934-10 51.2
1934-11 42.8
1934-12 45.8
1935-01 40.4
1935-02 42.6
1935-03 43.5
1935-04 47.1
1935-05 50.0
1935-06 60.5
1935-07 64.6
1935-08 64.0
1935-09 56.8
1935-10 48.6
1935-11 44.2
1935-12 36.4
1936-01 37.3
1936-02 35.0
1936-03 44.0
1936-04 43.9
1936-05 52.7
1936-06 58.6
1936-07 60.0
1936-08 61.1
1936-09 58.1
1936-10 49.6
1936-11 41.6
1936-12 41.3
1937-01 40.8
1937-02 41.0
1937-03 38.4
1937-04 47.4
1937-05 54.1
1937-06 58.6
1937-07 61.4
1937-08 61.8
1937-09 56.3
1937-10 50.9
1937-11 41.4
1937-12 37.1
1938-01 42.1
1938-02 41.2
1938-03 47.3
1938-04 46.6
1938-05 52.4
1938-06 59.0
1938-07 59.6
1938-08 60.4
1938-09 57.0
1938-10 50.7
1938-11 47.8
1938-12 39.2
1939-01 39.4
1939-02 40.9
1939-03 42.4
1939-04 47.8
1939-05 52.4
1939-06 58.0
1939-07 60.7
1939-08 61.8
1939-09 58.2
1939-10 46.7
1939-11 46.6
1939-12 37.8
Raw
logicbelief-master-puwork_lstm_chapter_12_monthly-shampoo-sales.csv
1-01 266.0
1-02 145.9
1-03 183.1
1-04 119.3
1-05 180.3
1-06 168.5
1-07 231.8
1-08 224.5
1-09 192.8
1-10 122.9
1-11 336.5
1-12 185.9
2-01 194.3
2-02 149.5
2-03 210.1
2-04 273.3
2-05 191.4
2-06 287.0
2-07 226.0
2-08 303.6
2-09 289.9
2-10 421.6
2-11 264.5
2-12 342.3
3-01 339.7
3-02 440.4
3-03 315.9
3-04 439.3
3-05 401.3
3-06 437.4
3-07 575.5
3-08 407.6
3-09 682.0
3-10 475.3
3-11 581.3
3-12 646.9
Raw
logicbelief-master-puwork_lstm_chapter_13_01_grid_search.py
# grid search sarima hyperparameters
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.statespace.sarimax import SARIMAX
from sklearn.metrics import mean_squared_error
# one-step sarima forecast
def sarima_forecast(history, config):
order, sorder, trend = config
# define model
model = SARIMAX(history, order=order, seasonal_order=sorder, trend=trend, enforce_stationarity=False, enforce_invertibility=False)
# fit model
model_fit = model.fit(disp=False)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = sarima_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of sarima configs to try
def sarima_configs(seasonal=[0]):
models = list()
# define config lists
p_params = [0, 1, 2]
d_params = [0, 1]
q_params = [0, 1, 2]
t_params = ['n','c','t','ct']
P_params = [0, 1, 2]
D_params = [0, 1]
Q_params = [0, 1, 2]
m_params = seasonal
# create config instances
for p in p_params:
for d in d_params:
for q in q_params:
for t in t_params:
for P in P_params:
for D in D_params:
for Q in Q_params:
for m in m_params:
cfg = [(p,d,q), (P,D,Q,m), t]
models.append(cfg)
return models
if __name__ == '__main__':
# define dataset
data = [10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0, 100.0]
print(data)
# data split
n_test = 4
# model configs
cfg_list = sarima_configs()
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_13_02_grid_search_daily_births.py
# grid search sarima hyperparameters for daily female dataset
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.statespace.sarimax import SARIMAX
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step sarima forecast
def sarima_forecast(history, config):
order, sorder, trend = config
# define model
model = SARIMAX(history, order=order, seasonal_order=sorder, trend=trend, enforce_stationarity=False, enforce_invertibility=False)
# fit model
model_fit = model.fit(disp=False)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = sarima_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of sarima configs to try
def sarima_configs(seasonal=[0]):
models = list()
# define config lists
p_params = [0, 1, 2]
d_params = [0, 1]
q_params = [0, 1, 2]
t_params = ['n','c','t','ct']
P_params = [0, 1, 2]
D_params = [0, 1]
Q_params = [0, 1, 2]
m_params = seasonal
# create config instances
for p in p_params:
for d in d_params:
for q in q_params:
for t in t_params:
for P in P_params:
for D in D_params:
for Q in Q_params:
for m in m_params:
cfg = [(p,d,q), (P,D,Q,m), t]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('daily-total-female-births.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 165
# model configs
cfg_list = sarima_configs()
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_13_03_grid_search_monthly_shampoo_sales.py
# grid search sarima hyperparameters for monthly shampoo sales dataset
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.statespace.sarimax import SARIMAX
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step sarima forecast
def sarima_forecast(history, config):
order, sorder, trend = config
# define model
model = SARIMAX(history, order=order, seasonal_order=sorder, trend=trend, enforce_stationarity=False, enforce_invertibility=False)
# fit model
model_fit = model.fit(disp=False)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = sarima_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of sarima configs to try
def sarima_configs(seasonal=[0]):
models = list()
# define config lists
p_params = [0, 1, 2]
d_params = [0, 1]
q_params = [0, 1, 2]
t_params = ['n','c','t','ct']
P_params = [0, 1, 2]
D_params = [0, 1]
Q_params = [0, 1, 2]
m_params = seasonal
# create config instances
for p in p_params:
for d in d_params:
for q in q_params:
for t in t_params:
for P in P_params:
for D in D_params:
for Q in Q_params:
for m in m_params:
cfg = [(p,d,q), (P,D,Q,m), t]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('monthly-shampoo-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = sarima_configs()
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_13_04_grid_search_monthly_mean_temp.py
# grid search sarima hyperparameters for monthly mean temp dataset
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.statespace.sarimax import SARIMAX
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step sarima forecast
def sarima_forecast(history, config):
order, sorder, trend = config
# define model
model = SARIMAX(history, order=order, seasonal_order=sorder, trend=trend, enforce_stationarity=False, enforce_invertibility=False)
# fit model
model_fit = model.fit(disp=False)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = sarima_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of sarima configs to try
def sarima_configs(seasonal=[0]):
models = list()
# define config lists
p_params = [0, 1, 2]
d_params = [0, 1]
q_params = [0, 1, 2]
t_params = ['n','c','t','ct']
P_params = [0, 1, 2]
D_params = [0, 1]
Q_params = [0, 1, 2]
m_params = seasonal
# create config instances
for p in p_params:
for d in d_params:
for q in q_params:
for t in t_params:
for P in P_params:
for D in D_params:
for Q in Q_params:
for m in m_params:
cfg = [(p,d,q), (P,D,Q,m), t]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('monthly-mean-temp.csv', header=0, index_col=0)
data = series.values
# trim dataset to 5 years
data = data[-(5*12):]
# data split
n_test = 12
# model configs
cfg_list = sarima_configs(seasonal=[0, 12])
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_13_05_grid_search_monthly_car_sales.py
# grid search sarima hyperparameters for monthly car sales dataset
from math import sqrt
from multiprocessing import cpu_count
from joblib import Parallel
from joblib import delayed
from warnings import catch_warnings
from warnings import filterwarnings
from statsmodels.tsa.statespace.sarimax import SARIMAX
from sklearn.metrics import mean_squared_error
from pandas import read_csv
# one-step sarima forecast
def sarima_forecast(history, config):
order, sorder, trend = config
# define model
model = SARIMAX(history, order=order, seasonal_order=sorder, trend=trend, enforce_stationarity=False, enforce_invertibility=False)
# fit model
model_fit = model.fit(disp=False)
# make one step forecast
yhat = model_fit.predict(len(history), len(history))
return yhat[0]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = sarima_forecast(history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
return error
# score a model, return None on failure
def score_model(data, n_test, cfg, debug=False):
result = None
# convert config to a key
key = str(cfg)
# show all warnings and fail on exception if debugging
if debug:
result = walk_forward_validation(data, n_test, cfg)
else:
# one failure during model validation suggests an unstable config
try:
# never show warnings when grid searching, too noisy
with catch_warnings():
filterwarnings("ignore")
result = walk_forward_validation(data, n_test, cfg)
except:
error = None
# check for an interesting result
if result is not None:
print(' > Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test, parallel=True):
scores = None
if parallel:
# execute configs in parallel
executor = Parallel(n_jobs=cpu_count(), backend='multiprocessing')
tasks = (delayed(score_model)(data, n_test, cfg) for cfg in cfg_list)
scores = executor(tasks)
else:
scores = [score_model(data, n_test, cfg) for cfg in cfg_list]
# remove empty results
scores = [r for r in scores if r[1] != None]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a set of sarima configs to try
def sarima_configs(seasonal=[0]):
models = list()
# define config lists
p_params = [0, 1, 2]
d_params = [0, 1]
q_params = [0, 1, 2]
t_params = ['n','c','t','ct']
P_params = [0, 1, 2]
D_params = [0, 1]
Q_params = [0, 1, 2]
m_params = seasonal
# create config instances
for p in p_params:
for d in d_params:
for q in q_params:
for t in t_params:
for P in P_params:
for D in D_params:
for Q in Q_params:
for m in m_params:
cfg = [(p,d,q), (P,D,Q,m), t]
models.append(cfg)
return models
if __name__ == '__main__':
# load dataset
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = sarima_configs(seasonal=[0,6,12])
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_13_daily-total-female-births.csv
1959-01-01 35
1959-01-02 32
1959-01-03 30
1959-01-04 31
1959-01-05 44
1959-01-06 29
1959-01-07 45
1959-01-08 43
1959-01-09 38
1959-01-10 27
1959-01-11 38
1959-01-12 33
1959-01-13 55
1959-01-14 47
1959-01-15 45
1959-01-16 37
1959-01-17 50
1959-01-18 43
1959-01-19 41
1959-01-20 52
1959-01-21 34
1959-01-22 53
1959-01-23 39
1959-01-24 32
1959-01-25 37
1959-01-26 43
1959-01-27 39
1959-01-28 35
1959-01-29 44
1959-01-30 38
1959-01-31 24
1959-02-01 23
1959-02-02 31
1959-02-03 44
1959-02-04 38
1959-02-05 50
1959-02-06 38
1959-02-07 51
1959-02-08 31
1959-02-09 31
1959-02-10 51
1959-02-11 36
1959-02-12 45
1959-02-13 51
1959-02-14 34
1959-02-15 52
1959-02-16 47
1959-02-17 45
1959-02-18 46
1959-02-19 39
1959-02-20 48
1959-02-21 37
1959-02-22 35
1959-02-23 52
1959-02-24 42
1959-02-25 45
1959-02-26 39
1959-02-27 37
1959-02-28 30
1959-03-01 35
1959-03-02 28
1959-03-03 45
1959-03-04 34
1959-03-05 36
1959-03-06 50
1959-03-07 44
1959-03-08 39
1959-03-09 32
1959-03-10 39
1959-03-11 45
1959-03-12 43
1959-03-13 39
1959-03-14 31
1959-03-15 27
1959-03-16 30
1959-03-17 42
1959-03-18 46
1959-03-19 41
1959-03-20 36
1959-03-21 45
1959-03-22 46
1959-03-23 43
1959-03-24 38
1959-03-25 34
1959-03-26 35
1959-03-27 56
1959-03-28 36
1959-03-29 32
1959-03-30 50
1959-03-31 41
1959-04-01 39
1959-04-02 41
1959-04-03 47
1959-04-04 34
1959-04-05 36
1959-04-06 33
1959-04-07 35
1959-04-08 38
1959-04-09 38
1959-04-10 34
1959-04-11 53
1959-04-12 34
1959-04-13 34
1959-04-14 38
1959-04-15 35
1959-04-16 32
1959-04-17 42
1959-04-18 34
1959-04-19 46
1959-04-20 30
1959-04-21 46
1959-04-22 45
1959-04-23 54
1959-04-24 34
1959-04-25 37
1959-04-26 35
1959-04-27 40
1959-04-28 42
1959-04-29 58
1959-04-30 51
1959-05-01 32
1959-05-02 35
1959-05-03 38
1959-05-04 33
1959-05-05 39
1959-05-06 47
1959-05-07 38
1959-05-08 52
1959-05-09 30
1959-05-10 34
1959-05-11 40
1959-05-12 35
1959-05-13 42
1959-05-14 41
1959-05-15 42
1959-05-16 38
1959-05-17 24
1959-05-18 34
1959-05-19 43
1959-05-20 36
1959-05-21 55
1959-05-22 41
1959-05-23 45
1959-05-24 41
1959-05-25 37
1959-05-26 43
1959-05-27 39
1959-05-28 33
1959-05-29 43
1959-05-30 40
1959-05-31 38
1959-06-01 45
1959-06-02 46
1959-06-03 34
1959-06-04 35
1959-06-05 48
1959-06-06 51
1959-06-07 36
1959-06-08 33
1959-06-09 46
1959-06-10 42
1959-06-11 48
1959-06-12 34
1959-06-13 41
1959-06-14 35
1959-06-15 40
1959-06-16 34
1959-06-17 30
1959-06-18 36
1959-06-19 40
1959-06-20 39
1959-06-21 45
1959-06-22 38
1959-06-23 47
1959-06-24 33
1959-06-25 30
1959-06-26 42
1959-06-27 43
1959-06-28 41
1959-06-29 41
1959-06-30 59
1959-07-01 43
1959-07-02 45
1959-07-03 38
1959-07-04 37
1959-07-05 45
1959-07-06 42
1959-07-07 57
1959-07-08 46
1959-07-09 51
1959-07-10 41
1959-07-11 47
1959-07-12 26
1959-07-13 35
1959-07-14 44
1959-07-15 41
1959-07-16 42
1959-07-17 36
1959-07-18 45
1959-07-19 45
1959-07-20 45
1959-07-21 47
1959-07-22 38
1959-07-23 42
1959-07-24 35
1959-07-25 36
1959-07-26 39
1959-07-27 45
1959-07-28 43
1959-07-29 47
1959-07-30 36
1959-07-31 41
1959-08-01 50
1959-08-02 39
1959-08-03 41
1959-08-04 46
1959-08-05 64
1959-08-06 45
1959-08-07 34
1959-08-08 38
1959-08-09 44
1959-08-10 48
1959-08-11 46
1959-08-12 44
1959-08-13 37
1959-08-14 39
1959-08-15 44
1959-08-16 45
1959-08-17 33
1959-08-18 44
1959-08-19 38
1959-08-20 46
1959-08-21 46
1959-08-22 40
1959-08-23 39
1959-08-24 44
1959-08-25 48
1959-08-26 50
1959-08-27 41
1959-08-28 42
1959-08-29 51
1959-08-30 41
1959-08-31 44
1959-09-01 38
1959-09-02 68
1959-09-03 40
1959-09-04 42
1959-09-05 51
1959-09-06 44
1959-09-07 45
1959-09-08 36
1959-09-09 57
1959-09-10 44
1959-09-11 42
1959-09-12 53
1959-09-13 42
1959-09-14 34
1959-09-15 40
1959-09-16 56
1959-09-17 44
1959-09-18 53
1959-09-19 55
1959-09-20 39
1959-09-21 59
1959-09-22 55
1959-09-23 73
1959-09-24 55
1959-09-25 44
1959-09-26 43
1959-09-27 40
1959-09-28 47
1959-09-29 51
1959-09-30 56
1959-10-01 49
1959-10-02 54
1959-10-03 56
1959-10-04 47
1959-10-05 44
1959-10-06 43
1959-10-07 42
1959-10-08 45
1959-10-09 50
1959-10-10 48
1959-10-11 43
1959-10-12 40
1959-10-13 59
1959-10-14 41
1959-10-15 42
1959-10-16 51
1959-10-17 49
1959-10-18 45
1959-10-19 43
1959-10-20 42
1959-10-21 38
1959-10-22 47
1959-10-23 38
1959-10-24 36
1959-10-25 42
1959-10-26 35
1959-10-27 28
1959-10-28 44
1959-10-29 36
1959-10-30 45
1959-10-31 46
1959-11-01 48
1959-11-02 49
1959-11-03 43
1959-11-04 42
1959-11-05 59
1959-11-06 45
1959-11-07 52
1959-11-08 46
1959-11-09 42
1959-11-10 40
1959-11-11 40
1959-11-12 45
1959-11-13 35
1959-11-14 35
1959-11-15 40
1959-11-16 39
1959-11-17 33
1959-11-18 42
1959-11-19 47
1959-11-20 51
1959-11-21 44
1959-11-22 40
1959-11-23 57
1959-11-24 49
1959-11-25 45
1959-11-26 49
1959-11-27 51
1959-11-28 46
1959-11-29 44
1959-11-30 52
1959-12-01 45
1959-12-02 32
1959-12-03 46
1959-12-04 41
1959-12-05 34
1959-12-06 33
1959-12-07 36
1959-12-08 49
1959-12-09 43
1959-12-10 43
1959-12-11 34
1959-12-12 39
1959-12-13 35
1959-12-14 52
1959-12-15 47
1959-12-16 52
1959-12-17 39
1959-12-18 40
1959-12-19 42
1959-12-20 42
1959-12-21 53
1959-12-22 39
1959-12-23 40
1959-12-24 38
1959-12-25 44
1959-12-26 34
1959-12-27 37
1959-12-28 52
1959-12-29 48
1959-12-30 55
1959-12-31 50
Raw
logicbelief-master-puwork_lstm_chapter_13_monthly-car-sales.csv
1960-01 6550
1960-02 8728
1960-03 12026
1960-04 14395
1960-05 14587
1960-06 13791
1960-07 9498
1960-08 8251
1960-09 7049
1960-10 9545
1960-11 9364
1960-12 8456
1961-01 7237
1961-02 9374
1961-03 11837
1961-04 13784
1961-05 15926
1961-06 13821
1961-07 11143
1961-08 7975
1961-09 7610
1961-10 10015
1961-11 12759
1961-12 8816
1962-01 10677
1962-02 10947
1962-03 15200
1962-04 17010
1962-05 20900
1962-06 16205
1962-07 12143
1962-08 8997
1962-09 5568
1962-10 11474
1962-11 12256
1962-12 10583
1963-01 10862
1963-02 10965
1963-03 14405
1963-04 20379
1963-05 20128
1963-06 17816
1963-07 12268
1963-08 8642
1963-09 7962
1963-10 13932
1963-11 15936
1963-12 12628
1964-01 12267
1964-02 12470
1964-03 18944
1964-04 21259
1964-05 22015
1964-06 18581
1964-07 15175
1964-08 10306
1964-09 10792
1964-10 14752
1964-11 13754
1964-12 11738
1965-01 12181
1965-02 12965
1965-03 19990
1965-04 23125
1965-05 23541
1965-06 21247
1965-07 15189
1965-08 14767
1965-09 10895
1965-10 17130
1965-11 17697
1965-12 16611
1966-01 12674
1966-02 12760
1966-03 20249
1966-04 22135
1966-05 20677
1966-06 19933
1966-07 15388
1966-08 15113
1966-09 13401
1966-10 16135
1966-11 17562
1966-12 14720
1967-01 12225
1967-02 11608
1967-03 20985
1967-04 19692
1967-05 24081
1967-06 22114
1967-07 14220
1967-08 13434
1967-09 13598
1967-10 17187
1967-11 16119
1967-12 13713
1968-01 13210
1968-02 14251
1968-03 20139
1968-04 21725
1968-05 26099
1968-06 21084
1968-07 18024
1968-08 16722
1968-09 14385
1968-10 21342
1968-11 17180
1968-12 14577
Raw
logicbelief-master-puwork_lstm_chapter_13_monthly-mean-temp.csv
1920-01 40.6
1920-02 40.8
1920-03 44.4
1920-04 46.7
1920-05 54.1
1920-06 58.5
1920-07 57.7
1920-08 56.4
1920-09 54.3
1920-10 50.5
1920-11 42.9
1920-12 39.8
1921-01 44.2
1921-02 39.8
1921-03 45.1
1921-04 47.0
1921-05 54.1
1921-06 58.7
1921-07 66.3
1921-08 59.9
1921-09 57.0
1921-10 54.2
1921-11 39.7
1921-12 42.8
1922-01 37.5
1922-02 38.7
1922-03 39.5
1922-04 42.1
1922-05 55.7
1922-06 57.8
1922-07 56.8
1922-08 54.3
1922-09 54.3
1922-10 47.1
1922-11 41.8
1922-12 41.7
1923-01 41.8
1923-02 40.1
1923-03 42.9
1923-04 45.8
1923-05 49.2
1923-06 52.7
1923-07 64.2
1923-08 59.6
1923-09 54.4
1923-10 49.2
1923-11 36.6
1923-12 37.6
1924-01 39.3
1924-02 37.5
1924-03 38.3
1924-04 45.5
1924-05 53.2
1924-06 57.7
1924-07 60.8
1924-08 58.2
1924-09 56.4
1924-10 49.8
1924-11 44.4
1924-12 43.6
1925-01 40.0
1925-02 40.5
1925-03 40.8
1925-04 45.1
1925-05 53.8
1925-06 59.4
1925-07 63.5
1925-08 61.0
1925-09 53.0
1925-10 50.0
1925-11 38.1
1925-12 36.3
1926-01 39.2
1926-02 43.4
1926-03 43.4
1926-04 48.9
1926-05 50.6
1926-06 56.8
1926-07 62.5
1926-08 62.0
1926-09 57.5
1926-10 46.7
1926-11 41.6
1926-12 39.8
1927-01 39.4
1927-02 38.5
1927-03 45.3
1927-04 47.1
1927-05 51.7
1927-06 55.0
1927-07 60.4
1927-08 60.5
1927-09 54.7
1927-10 50.3
1927-11 42.3
1927-12 35.2
1928-01 40.8
1928-02 41.1
1928-03 42.8
1928-04 47.3
1928-05 50.9
1928-06 56.4
1928-07 62.2
1928-08 60.5
1928-09 55.4
1928-10 50.2
1928-11 43.0
1928-12 37.3
1929-01 34.8
1929-02 31.3
1929-03 41.0
1929-04 43.9
1929-05 53.1
1929-06 56.9
1929-07 62.5
1929-08 60.3
1929-09 59.8
1929-10 49.2
1929-11 42.9
1929-12 41.9
1930-01 41.6
1930-02 37.1
1930-03 41.2
1930-04 46.9
1930-05 51.2
1930-06 60.4
1930-07 60.1
1930-08 61.6
1930-09 57.0
1930-10 50.9
1930-11 43.0
1930-12 38.8
1931-01 37.1
1931-02 38.4
1931-03 38.4
1931-04 46.5
1931-05 53.5
1931-06 58.4
1931-07 60.6
1931-08 58.2
1931-09 53.8
1931-10 46.6
1931-11 45.5
1931-12 40.6
1932-01 42.4
1932-02 38.4
1932-03 40.3
1932-04 44.6
1932-05 50.9
1932-06 57.0
1932-07 62.1
1932-08 63.5
1932-09 56.2
1932-10 47.3
1932-11 43.6
1932-12 41.8
1933-01 36.2
1933-02 39.3
1933-03 44.5
1933-04 48.7
1933-05 54.2
1933-06 60.8
1933-07 65.5
1933-08 64.9
1933-09 60.1
1933-10 50.2
1933-11 42.1
1933-12 35.6
1934-01 39.4
1934-02 38.2
1934-03 40.4
1934-04 46.9
1934-05 53.4
1934-06 59.6
1934-07 66.5
1934-08 60.4
1934-09 59.2
1934-10 51.2
1934-11 42.8
1934-12 45.8
1935-01 40.4
1935-02 42.6
1935-03 43.5
1935-04 47.1
1935-05 50.0
1935-06 60.5
1935-07 64.6
1935-08 64.0
1935-09 56.8
1935-10 48.6
1935-11 44.2
1935-12 36.4
1936-01 37.3
1936-02 35.0
1936-03 44.0
1936-04 43.9
1936-05 52.7
1936-06 58.6
1936-07 60.0
1936-08 61.1
1936-09 58.1
1936-10 49.6
1936-11 41.6
1936-12 41.3
1937-01 40.8
1937-02 41.0
1937-03 38.4
1937-04 47.4
1937-05 54.1
1937-06 58.6
1937-07 61.4
1937-08 61.8
1937-09 56.3
1937-10 50.9
1937-11 41.4
1937-12 37.1
1938-01 42.1
1938-02 41.2
1938-03 47.3
1938-04 46.6
1938-05 52.4
1938-06 59.0
1938-07 59.6
1938-08 60.4
1938-09 57.0
1938-10 50.7
1938-11 47.8
1938-12 39.2
1939-01 39.4
1939-02 40.9
1939-03 42.4
1939-04 47.8
1939-05 52.4
1939-06 58.0
1939-07 60.7
1939-08 61.8
1939-09 58.2
1939-10 46.7
1939-11 46.6
1939-12 37.8
Raw
logicbelief-master-puwork_lstm_chapter_13_monthly-shampoo-sales.csv
1-01 266.0
1-02 145.9
1-03 183.1
1-04 119.3
1-05 180.3
1-06 168.5
1-07 231.8
1-08 224.5
1-09 192.8
1-10 122.9
1-11 336.5
1-12 185.9
2-01 194.3
2-02 149.5
2-03 210.1
2-04 273.3
2-05 191.4
2-06 287.0
2-07 226.0
2-08 303.6
2-09 289.9
2-10 421.6
2-11 264.5
2-12 342.3
3-01 339.7
3-02 440.4
3-03 315.9
3-04 439.3
3-05 401.3
3-06 437.4
3-07 575.5
3-08 407.6
3-09 682.0
3-10 475.3
3-11 581.3
3-12 646.9
Raw
logicbelief-master-puwork_lstm_chapter_14_01_simple_forecast.py
# persistence forecast for monthly car sales dataset
from math import sqrt
from numpy import median
from numpy import mean
from numpy import std
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# difference dataset
def difference(data, interval):
return [data[i] - data[i - interval] for i in range(interval, len(data))]
# fit a model
def model_fit(train, config):
return None
# forecast with a pre-fit model
def model_predict(model, history, config):
values = list()
for offset in config:
values.append(history[-offset])
return median(values)
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# repeat evaluation of a config
def repeat_evaluate(data, config, n_test, n_repeats=30):
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
return scores
# summarize model performance
def summarize_scores(name, scores):
# print a summary
scores_m, score_std = mean(scores), std(scores)
print('%s: %.3f RMSE (+/- %.3f)' % (name, scores_m, score_std))
# box and whisker plot
pyplot.boxplot(scores)
pyplot.show()
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# define config
config = [12, 24, 36]
# grid search
scores = repeat_evaluate(data, config, n_test)
# summarize scores
summarize_scores('persistence', scores)
Raw
logicbelief-master-puwork_lstm_chapter_14_02_mlp_forecast_model.py
# evaluate mlp for monthly car sales dataset
from math import sqrt
from numpy import array
from numpy import mean
from numpy import std
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from matplotlib import pyplot
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# fit a model
def model_fit(train, config):
# unpack config
n_input, n_nodes, n_epochs, n_batch = config
# prepare data
data = series_to_supervised(train, n_input)
train_x, train_y = data[:, :-1], data[:, -1]
# define model
model = Sequential()
model.add(Dense(n_nodes, activation='relu', input_dim=n_input))
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with a pre-fit model
def model_predict(model, history, config):
# unpack config
n_input, _, _, _ = config
# prepare data
x_input = array(history[-n_input:]).reshape(1, n_input)
# forecast
yhat = model.predict(x_input, verbose=0)
return yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# repeat evaluation of a config
def repeat_evaluate(data, config, n_test, n_repeats=30):
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
return scores
# summarize model performance
def summarize_scores(name, scores):
# print a summary
scores_m, score_std = mean(scores), std(scores)
print('%s: %.3f RMSE (+/- %.3f)' % (name, scores_m, score_std))
# box and whisker plot
pyplot.boxplot(scores)
pyplot.show()
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# define config
config = [24, 500, 100, 100]
# grid search
scores = repeat_evaluate(data, config, n_test)
# summarize scores
summarize_scores('mlp', scores)
Raw
logicbelief-master-puwork_lstm_chapter_14_03_cnn_forecast_model.py
# evaluate cnn for monthly car sales dataset
from math import sqrt
from numpy import array
from numpy import mean
from numpy import std
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from matplotlib import pyplot
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# fit a model
def model_fit(train, config):
# unpack config
n_input, n_filters, n_kernel, n_epochs, n_batch = config
# prepare data
data = series_to_supervised(train, n_input)
train_x, train_y = data[:, :-1], data[:, -1]
train_x = train_x.reshape((train_x.shape[0], train_x.shape[1], 1))
# define model
model = Sequential()
model.add(Conv1D(filters=n_filters, kernel_size=n_kernel, activation='relu', input_shape=(n_input, 1)))
model.add(Conv1D(filters=n_filters, kernel_size=n_kernel, activation='relu'))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with a pre-fit model
def model_predict(model, history, config):
# unpack config
n_input, _, _, _, _ = config
# prepare data
x_input = array(history[-n_input:]).reshape((1, n_input, 1))
# forecast
yhat = model.predict(x_input, verbose=0)
return yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# repeat evaluation of a config
def repeat_evaluate(data, config, n_test, n_repeats=30):
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
return scores
# summarize model performance
def summarize_scores(name, scores):
# print a summary
scores_m, score_std = mean(scores), std(scores)
print('%s: %.3f RMSE (+/- %.3f)' % (name, scores_m, score_std))
# box and whisker plot
pyplot.boxplot(scores)
pyplot.show()
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# define config
config = [36, 256, 3, 100, 100]
# grid search
scores = repeat_evaluate(data, config, n_test)
# summarize scores
summarize_scores('cnn', scores)
Raw
logicbelief-master-puwork_lstm_chapter_14_04_lstm_forecast_model.py
# evaluate lstm for monthly car sales dataset
from math import sqrt
from numpy import array
from numpy import mean
from numpy import std
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from matplotlib import pyplot
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# difference dataset
def difference(data, interval):
return [data[i] - data[i - interval] for i in range(interval, len(data))]
# fit a model
def model_fit(train, config):
# unpack config
n_input, n_nodes, n_epochs, n_batch, n_diff = config
# prepare data
if n_diff > 0:
train = difference(train, n_diff)
data = series_to_supervised(train, n_input)
train_x, train_y = data[:, :-1], data[:, -1]
train_x = train_x.reshape((train_x.shape[0], train_x.shape[1], 1))
# define model
model = Sequential()
model.add(LSTM(n_nodes, activation='relu', input_shape=(n_input, 1)))
model.add(Dense(n_nodes, activation='relu'))
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with a pre-fit model
def model_predict(model, history, config):
# unpack config
n_input, _, _, _, n_diff = config
# prepare data
correction = 0.0
if n_diff > 0:
correction = history[-n_diff]
history = difference(history, n_diff)
x_input = array(history[-n_input:]).reshape((1, n_input, 1))
# forecast
yhat = model.predict(x_input, verbose=0)
return correction + yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# repeat evaluation of a config
def repeat_evaluate(data, config, n_test, n_repeats=30):
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
return scores
# summarize model performance
def summarize_scores(name, scores):
# print a summary
scores_m, score_std = mean(scores), std(scores)
print('%s: %.3f RMSE (+/- %.3f)' % (name, scores_m, score_std))
# box and whisker plot
pyplot.boxplot(scores)
pyplot.show()
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# define config
config = [36, 50, 100, 100, 12]
# grid search
scores = repeat_evaluate(data, config, n_test)
# summarize scores
summarize_scores('lstm', scores)
Raw
logicbelief-master-puwork_lstm_chapter_14_05_cnn_lstm_forecast_model.py
# evaluate cnn-lstm for monthly car sales dataset
from math import sqrt
from numpy import array
from numpy import mean
from numpy import std
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from keras.layers import TimeDistributed
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from matplotlib import pyplot
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# fit a model
def model_fit(train, config):
# unpack config
n_seq, n_steps, n_filters, n_kernel, n_nodes, n_epochs, n_batch = config
n_input = n_seq * n_steps
# prepare data
data = series_to_supervised(train, n_input)
train_x, train_y = data[:, :-1], data[:, -1]
train_x = train_x.reshape((train_x.shape[0], n_seq, n_steps, 1))
# define model
model = Sequential()
model.add(TimeDistributed(Conv1D(filters=n_filters, kernel_size=n_kernel, activation='relu', input_shape=(None,n_steps,1))))
model.add(TimeDistributed(Conv1D(filters=n_filters, kernel_size=n_kernel, activation='relu')))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(n_nodes, activation='relu'))
model.add(Dense(n_nodes, activation='relu'))
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with a pre-fit model
def model_predict(model, history, config):
# unpack config
n_seq, n_steps, _, _, _, _, _ = config
n_input = n_seq * n_steps
# prepare data
x_input = array(history[-n_input:]).reshape((1, n_seq, n_steps, 1))
# forecast
yhat = model.predict(x_input, verbose=0)
return yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# repeat evaluation of a config
def repeat_evaluate(data, config, n_test, n_repeats=30):
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
return scores
# summarize model performance
def summarize_scores(name, scores):
# print a summary
scores_m, score_std = mean(scores), std(scores)
print('%s: %.3f RMSE (+/- %.3f)' % (name, scores_m, score_std))
# box and whisker plot
pyplot.boxplot(scores)
pyplot.show()
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# define config
config = [3, 12, 64, 3, 100, 200, 100]
# grid search
scores = repeat_evaluate(data, config, n_test)
# summarize scores
summarize_scores('cnn-lstm', scores)
Raw
logicbelief-master-puwork_lstm_chapter_14_06_convlstm_forecast_model.py
# evaluate convlstm for monthly car sales dataset
from math import sqrt
from numpy import array
from numpy import mean
from numpy import std
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import ConvLSTM2D
from matplotlib import pyplot
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# difference dataset
def difference(data, interval):
return [data[i] - data[i - interval] for i in range(interval, len(data))]
# fit a model
def model_fit(train, config):
# unpack config
n_seq, n_steps, n_filters, n_kernel, n_nodes, n_epochs, n_batch = config
n_input = n_seq * n_steps
# prepare data
data = series_to_supervised(train, n_input)
train_x, train_y = data[:, :-1], data[:, -1]
train_x = train_x.reshape((train_x.shape[0], n_seq, 1, n_steps, 1))
# define model
model = Sequential()
model.add(ConvLSTM2D(filters=n_filters, kernel_size=(1,n_kernel), activation='relu', input_shape=(n_seq, 1, n_steps, 1)))
model.add(Flatten())
model.add(Dense(n_nodes, activation='relu'))
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with a pre-fit model
def model_predict(model, history, config):
# unpack config
n_seq, n_steps, _, _, _, _, _ = config
n_input = n_seq * n_steps
# prepare data
x_input = array(history[-n_input:]).reshape((1, n_seq, 1, n_steps, 1))
# forecast
yhat = model.predict(x_input, verbose=0)
return yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# repeat evaluation of a config
def repeat_evaluate(data, config, n_test, n_repeats=30):
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
return scores
# summarize model performance
def summarize_scores(name, scores):
# print a summary
scores_m, score_std = mean(scores), std(scores)
print('%s: %.3f RMSE (+/- %.3f)' % (name, scores_m, score_std))
# box and whisker plot
pyplot.boxplot(scores)
pyplot.show()
series = read_csv('monthly-car-sales.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# define config
config = [3, 12, 256, 3, 200, 200, 100]
# grid search
scores = repeat_evaluate(data, config, n_test)
# summarize scores
summarize_scores('convlstm', scores)
Raw
logicbelief-master-puwork_lstm_chapter_14_monthly-car-sales.csv
1960-01 6550
1960-02 8728
1960-03 12026
1960-04 14395
1960-05 14587
1960-06 13791
1960-07 9498
1960-08 8251
1960-09 7049
1960-10 9545
1960-11 9364
1960-12 8456
1961-01 7237
1961-02 9374
1961-03 11837
1961-04 13784
1961-05 15926
1961-06 13821
1961-07 11143
1961-08 7975
1961-09 7610
1961-10 10015
1961-11 12759
1961-12 8816
1962-01 10677
1962-02 10947
1962-03 15200
1962-04 17010
1962-05 20900
1962-06 16205
1962-07 12143
1962-08 8997
1962-09 5568
1962-10 11474
1962-11 12256
1962-12 10583
1963-01 10862
1963-02 10965
1963-03 14405
1963-04 20379
1963-05 20128
1963-06 17816
1963-07 12268
1963-08 8642
1963-09 7962
1963-10 13932
1963-11 15936
1963-12 12628
1964-01 12267
1964-02 12470
1964-03 18944
1964-04 21259
1964-05 22015
1964-06 18581
1964-07 15175
1964-08 10306
1964-09 10792
1964-10 14752
1964-11 13754
1964-12 11738
1965-01 12181
1965-02 12965
1965-03 19990
1965-04 23125
1965-05 23541
1965-06 21247
1965-07 15189
1965-08 14767
1965-09 10895
1965-10 17130
1965-11 17697
1965-12 16611
1966-01 12674
1966-02 12760
1966-03 20249
1966-04 22135
1966-05 20677
1966-06 19933
1966-07 15388
1966-08 15113
1966-09 13401
1966-10 16135
1966-11 17562
1966-12 14720
1967-01 12225
1967-02 11608
1967-03 20985
1967-04 19692
1967-05 24081
1967-06 22114
1967-07 14220
1967-08 13434
1967-09 13598
1967-10 17187
1967-11 16119
1967-12 13713
1968-01 13210
1968-02 14251
1968-03 20139
1968-04 21725
1968-05 26099
1968-06 21084
1968-07 18024
1968-08 16722
1968-09 14385
1968-10 21342
1968-11 17180
1968-12 14577
Raw
logicbelief-master-puwork_lstm_chapter_15_01_load_dataset.py
# load and plot monthly airline passengers dataset
from pandas import read_csv
from matplotlib import pyplot
# load
series = read_csv('monthly-airline-passengers.csv', header=0, index_col=0)
# summarize shape
print(series.shape)
# plot
pyplot.plot(series)
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_15_02_grid_search_persistence.py
# grid search persistence models for monthly airline passengers dataset
from math import sqrt
from numpy import mean
from pandas import read_csv
from sklearn.metrics import mean_squared_error
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# fit a model
def model_fit(train, config):
return None
# forecast with a pre-fit model
def model_predict(model, history, offset):
return history[-offset]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# score a model, return None on failure
def repeat_evaluate(data, config, n_test, n_repeats=10):
# convert config to a key
key = str(config)
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
# summarize score
result = mean(scores)
print('> Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test):
# evaluate configs
scores = scores = [repeat_evaluate(data, cfg, n_test) for cfg in cfg_list]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# define dataset
series = read_csv('monthly-airline-passengers.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = [1, 6, 12, 24, 36]
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 10 configs
for cfg, error in scores[:10]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_15_03_grid_search_mlp.py
# grid search mlps for monthly airline passengers dataset
from math import sqrt
from numpy import array
from numpy import mean
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# difference dataset
def difference(data, order):
return [data[i] - data[i - order] for i in range(order, len(data))]
# fit a model
def model_fit(train, config):
# unpack config
n_input, n_nodes, n_epochs, n_batch, n_diff = config
# prepare data
if n_diff > 0:
train = difference(train, n_diff)
# transform series into supervised format
data = series_to_supervised(train, n_in=n_input)
# separate inputs and outputs
train_x, train_y = data[:, :-1], data[:, -1]
# define model
model = Sequential()
model.add(Dense(n_nodes, activation='relu', input_dim=n_input))
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit model
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with the fit model
def model_predict(model, history, config):
# unpack config
n_input, _, _, _, n_diff = config
# prepare data
correction = 0.0
if n_diff > 0:
correction = history[-n_diff]
history = difference(history, n_diff)
# shape input for model
x_input = array(history[-n_input:]).reshape((1, n_input))
# make forecast
yhat = model.predict(x_input, verbose=0)
# correct forecast if it was differenced
return correction + yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# score a model, return None on failure
def repeat_evaluate(data, config, n_test, n_repeats=10):
# convert config to a key
key = str(config)
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
# summarize score
result = mean(scores)
print('> Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test):
# evaluate configs
scores = scores = [repeat_evaluate(data, cfg, n_test) for cfg in cfg_list]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a list of configs to try
def model_configs():
# define scope of configs
n_input = [12]
n_nodes = [50, 100]
n_epochs = [100]
n_batch = [1, 150]
n_diff = [0, 12]
# create configs
configs = list()
for i in n_input:
for j in n_nodes:
for k in n_epochs:
for l in n_batch:
for m in n_diff:
cfg = [i, j, k, l, m]
configs.append(cfg)
print('Total configs: %d' % len(configs))
return configs
# define dataset
series = read_csv('monthly-airline-passengers.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = model_configs()
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 3 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_15_04_grid_search_cnn.py
# grid search cnn for monthly airline passengers dataset
from math import sqrt
from numpy import array
from numpy import mean
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# difference dataset
def difference(data, order):
return [data[i] - data[i - order] for i in range(order, len(data))]
# fit a model
def model_fit(train, config):
# unpack config
n_input, n_filters, n_kernel, n_epochs, n_batch, n_diff = config
# prepare data
if n_diff > 0:
train = difference(train, n_diff)
# transform series into supervised format
data = series_to_supervised(train, n_in=n_input)
# separate inputs and outputs
train_x, train_y = data[:, :-1], data[:, -1]
# reshape input data into [samples, timesteps, features]
n_features = 1
train_x = train_x.reshape((train_x.shape[0], train_x.shape[1], n_features))
# define model
model = Sequential()
model.add(Conv1D(filters=n_filters, kernel_size=n_kernel, activation='relu', input_shape=(n_input, n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with the fit model
def model_predict(model, history, config):
# unpack config
n_input, _, _, _, _, n_diff = config
# prepare data
correction = 0.0
if n_diff > 0:
correction = history[-n_diff]
history = difference(history, n_diff)
x_input = array(history[-n_input:]).reshape((1, n_input, 1))
# forecast
yhat = model.predict(x_input, verbose=0)
return correction + yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# score a model, return None on failure
def repeat_evaluate(data, config, n_test, n_repeats=10):
# convert config to a key
key = str(config)
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
# summarize score
result = mean(scores)
print('> Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test):
# evaluate configs
scores = scores = [repeat_evaluate(data, cfg, n_test) for cfg in cfg_list]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a list of configs to try
def model_configs():
# define scope of configs
n_input = [12]
n_filters = [64]
n_kernels = [3, 5]
n_epochs = [100]
n_batch = [1, 150]
n_diff = [0, 12]
# create configs
configs = list()
for a in n_input:
for b in n_filters:
for c in n_kernels:
for d in n_epochs:
for e in n_batch:
for f in n_diff:
cfg = [a,b,c,d,e,f]
configs.append(cfg)
print('Total configs: %d' % len(configs))
return configs
# define dataset
series = read_csv('monthly-airline-passengers.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = model_configs()
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 10 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_15_05_grid_search_lstm.py
# grid search lstm for monthly airline passengers dataset
from math import sqrt
from numpy import array
from numpy import mean
from pandas import DataFrame
from pandas import concat
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
# split a univariate dataset into train/test sets
def train_test_split(data, n_test):
return data[:-n_test], data[-n_test:]
# transform list into supervised learning format
def series_to_supervised(data, n_in, n_out=1):
df = DataFrame(data)
cols = list()
# input sequence (t-n, ... t-1)
for i in range(n_in, 0, -1):
cols.append(df.shift(i))
# forecast sequence (t, t+1, ... t+n)
for i in range(0, n_out):
cols.append(df.shift(-i))
# put it all together
agg = concat(cols, axis=1)
# drop rows with NaN values
agg.dropna(inplace=True)
return agg.values
# root mean squared error or rmse
def measure_rmse(actual, predicted):
return sqrt(mean_squared_error(actual, predicted))
# difference dataset
def difference(data, order):
return [data[i] - data[i - order] for i in range(order, len(data))]
# fit a model
def model_fit(train, config):
# unpack config
n_input, n_nodes, n_epochs, n_batch, n_diff = config
# prepare data
if n_diff > 0:
train = difference(train, n_diff)
# transform series into supervised format
data = series_to_supervised(train, n_in=n_input)
# separate inputs and outputs
train_x, train_y = data[:, :-1], data[:, -1]
# reshape input data into [samples, timesteps, features]
n_features = 1
train_x = train_x.reshape((train_x.shape[0], train_x.shape[1], n_features))
# define model
model = Sequential()
model.add(LSTM(n_nodes, activation='relu', input_shape=(n_input, n_features)))
model.add(Dense(n_nodes, activation='relu'))
model.add(Dense(1))
model.compile(loss='mse', optimizer='adam')
# fit model
model.fit(train_x, train_y, epochs=n_epochs, batch_size=n_batch, verbose=0)
return model
# forecast with the fit model
def model_predict(model, history, config):
# unpack config
n_input, _, _, _, n_diff = config
# prepare data
correction = 0.0
if n_diff > 0:
correction = history[-n_diff]
history = difference(history, n_diff)
# reshape sample into [samples, timesteps, features]
x_input = array(history[-n_input:]).reshape((1, n_input, 1))
# forecast
yhat = model.predict(x_input, verbose=0)
return correction + yhat[0]
# walk-forward validation for univariate data
def walk_forward_validation(data, n_test, cfg):
predictions = list()
# split dataset
train, test = train_test_split(data, n_test)
# fit model
model = model_fit(train, cfg)
# seed history with training dataset
history = [x for x in train]
# step over each time-step in the test set
for i in range(len(test)):
# fit model and make forecast for history
yhat = model_predict(model, history, cfg)
# store forecast in list of predictions
predictions.append(yhat)
# add actual observation to history for the next loop
history.append(test[i])
# estimate prediction error
error = measure_rmse(test, predictions)
print(' > %.3f' % error)
return error
# score a model, return None on failure
def repeat_evaluate(data, config, n_test, n_repeats=10):
# convert config to a key
key = str(config)
# fit and evaluate the model n times
scores = [walk_forward_validation(data, n_test, config) for _ in range(n_repeats)]
# summarize score
result = mean(scores)
print('> Model[%s] %.3f' % (key, result))
return (key, result)
# grid search configs
def grid_search(data, cfg_list, n_test):
# evaluate configs
scores = scores = [repeat_evaluate(data, cfg, n_test) for cfg in cfg_list]
# sort configs by error, asc
scores.sort(key=lambda tup: tup[1])
return scores
# create a list of configs to try
def model_configs():
# define scope of configs
n_input = [12]
n_nodes = [100]
n_epochs = [50]
n_batch = [1, 150]
n_diff = [12]
# create configs
configs = list()
for i in n_input:
for j in n_nodes:
for k in n_epochs:
for l in n_batch:
for m in n_diff:
cfg = [i, j, k, l, m]
configs.append(cfg)
print('Total configs: %d' % len(configs))
return configs
# define dataset
series = read_csv('monthly-airline-passengers.csv', header=0, index_col=0)
data = series.values
# data split
n_test = 12
# model configs
cfg_list = model_configs()
# grid search
scores = grid_search(data, cfg_list, n_test)
print('done')
# list top 10 configs
for cfg, error in scores[:3]:
print(cfg, error)
Raw
logicbelief-master-puwork_lstm_chapter_15_monthly-airline-passengers.csv
Month Passengers
1949-01 112
1949-02 118
1949-03 132
1949-04 129
1949-05 121
1949-06 135
1949-07 148
1949-08 148
1949-09 136
1949-10 119
1949-11 104
1949-12 118
1950-01 115
1950-02 126
1950-03 141
1950-04 135
1950-05 125
1950-06 149
1950-07 170
1950-08 170
1950-09 158
1950-10 133
1950-11 114
1950-12 140
1951-01 145
1951-02 150
1951-03 178
1951-04 163
1951-05 172
1951-06 178
1951-07 199
1951-08 199
1951-09 184
1951-10 162
1951-11 146
1951-12 166
1952-01 171
1952-02 180
1952-03 193
1952-04 181
1952-05 183
1952-06 218
1952-07 230
1952-08 242
1952-09 209
1952-10 191
1952-11 172
1952-12 194
1953-01 196
1953-02 196
1953-03 236
1953-04 235
1953-05 229
1953-06 243
1953-07 264
1953-08 272
1953-09 237
1953-10 211
1953-11 180
1953-12 201
1954-01 204
1954-02 188
1954-03 235
1954-04 227
1954-05 234
1954-06 264
1954-07 302
1954-08 293
1954-09 259
1954-10 229
1954-11 203
1954-12 229
1955-01 242
1955-02 233
1955-03 267
1955-04 269
1955-05 270
1955-06 315
1955-07 364
1955-08 347
1955-09 312
1955-10 274
1955-11 237
1955-12 278
1956-01 284
1956-02 277
1956-03 317
1956-04 313
1956-05 318
1956-06 374
1956-07 413
1956-08 405
1956-09 355
1956-10 306
1956-11 271
1956-12 306
1957-01 315
1957-02 301
1957-03 356
1957-04 348
1957-05 355
1957-06 422
1957-07 465
1957-08 467
1957-09 404
1957-10 347
1957-11 305
1957-12 336
1958-01 340
1958-02 318
1958-03 362
1958-04 348
1958-05 363
1958-06 435
1958-07 491
1958-08 505
1958-09 404
1958-10 359
1958-11 310
1958-12 337
1959-01 360
1959-02 342
1959-03 406
1959-04 396
1959-05 420
1959-06 472
1959-07 548
1959-08 559
1959-09 463
1959-10 407
1959-11 362
1959-12 405
1960-01 417
1960-02 391
1960-03 419
1960-04 461
1960-05 472
1960-06 535
1960-07 622
1960-08 606
1960-09 508
1960-10 461
1960-11 390
1960-12 432
Raw
logicbelief-master-puwork_lstm_chapter_16_01_prepare_data.py
# load and clean-up power usage data
from numpy import nan
from pandas import read_csv
# load all data
dataset = read_csv('household_power_consumption.txt', sep=';', header=0, low_memory=False, infer_datetime_format=True, parse_dates={'datetime':[0,1]}, index_col=['datetime'])
# summarize
print(dataset.shape)
print(dataset.head())
# mark all missing values
dataset.replace('?', nan, inplace=True)
# add a column for for the remainder of sub metering
values = dataset.values.astype('float32')
dataset['sub_metering_4'] = (values[:,0] * 1000 / 60) - (values[:,4] + values[:,5] + values[:,6])
# save updated dataset
dataset.to_csv('household_power_consumption.csv')
# load the new dataset and summarize
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
print(dataset.head())
Raw
logicbelief-master-puwork_lstm_chapter_16_02_line_plots.py
# line plots for power usage dataset
from pandas import read_csv
from matplotlib import pyplot
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# line plot for each variable
pyplot.figure()
for i in range(len(dataset.columns)):
# create subplot
pyplot.subplot(len(dataset.columns), 1, i+1)
# get variable name
name = dataset.columns[i]
# plot data
pyplot.plot(dataset[name])
# set title
pyplot.title(name, y=0)
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_16_03_plots_of_yearly_power_consumption.py
# yearly line plots for power usage dataset
from pandas import read_csv
from matplotlib import pyplot
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# plot active power for each year
years = ['2007', '2008', '2009', '2010']
pyplot.figure()
for i in range(len(years)):
# prepare subplot
ax = pyplot.subplot(len(years), 1, i+1)
# determine the year to plot
year = years[i]
# get all observations for the year
result = dataset[str(year)]
# plot the active power for the year
pyplot.plot(result['Global_active_power'])
# add a title to the subplot
pyplot.title(str(year), y=0, loc='left')
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_16_04_plots_of_monthly_power_consumption.py
# monthly line plots for power usage dataset
from pandas import read_csv
from matplotlib import pyplot
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# plot active power for each year
months = [x for x in range(1, 13)]
pyplot.figure()
for i in range(len(months)):
# prepare subplot
ax = pyplot.subplot(len(months), 1, i+1)
# determine the month to plot
month = '2007-' + str(months[i])
# get all observations for the month
result = dataset[month]
# plot the active power for the month
pyplot.plot(result['Global_active_power'])
# add a title to the subplot
pyplot.title(month, y=0, loc='left')
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_16_05_plots_of_daily_power_consumption.py
# daily line plots for power usage dataset
from pandas import read_csv
from matplotlib import pyplot
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# plot active power for each year
days = [x for x in range(1, 20)]
pyplot.figure()
for i in range(len(days)):
# prepare subplot
ax = pyplot.subplot(len(days), 1, i+1)
# determine the day to plot
day = '2007-01-' + str(days[i])
# get all observations for the day
result = dataset[day]
# plot the active power for the day
pyplot.plot(result['Global_active_power'])
# add a title to the subplot
pyplot.title(day, y=0, loc='left', size=6)
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_16_06_histogram_all_variables.py
# histogram plots for power usage dataset
from pandas import read_csv
from matplotlib import pyplot
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# histogram plot for each variable
pyplot.figure()
for i in range(len(dataset.columns)):
# create subplot
pyplot.subplot(len(dataset.columns), 1, i+1)
# get variable name
name = dataset.columns[i]
# create histogram
dataset[name].hist(bins=100)
# set title
pyplot.title(name, y=0, loc='right')
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_16_07_histogram_yearly_power_consumption.py
# yearly histogram plots for power usage dataset
from pandas import read_csv
from matplotlib import pyplot
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# plot active power for each year
years = ['2007', '2008', '2009', '2010']
pyplot.figure()
for i in range(len(years)):
# prepare subplot
ax = pyplot.subplot(len(years), 1, i+1)
# determine the year to plot
year = years[i]
# get all observations for the year
result = dataset[str(year)]
# plot the active power for the year
result['Global_active_power'].hist(bins=100)
# zoom in on the distribution
ax.set_xlim(0, 5)
# add a title to the subplot
pyplot.title(str(year), y=0, loc='right')
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_16_08_histogram_monthly_power_consumption.py
# monthly histogram plots for power usage dataset
from pandas import read_csv
from matplotlib import pyplot
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# plot active power for each year
months = [x for x in range(1, 13)]
pyplot.figure()
for i in range(len(months)):
# prepare subplot
ax = pyplot.subplot(len(months), 1, i+1)
# determine the month to plot
month = '2007-' + str(months[i])
# get all observations for the month
result = dataset[month]
# plot the active power for the month
result['Global_active_power'].hist(bins=100)
# zoom in on the distribution
ax.set_xlim(0, 5)
# add a title to the subplot
pyplot.title(month, y=0, loc='right')
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_17_01_prepare_data.py
# load and clean-up the power usage dataset
from numpy import nan
from numpy import isnan
from pandas import read_csv
# fill missing values with a value at the same time one day ago
def fill_missing(values):
one_day = 60 * 24
for row in range(values.shape[0]):
for col in range(values.shape[1]):
if isnan(values[row, col]):
values[row, col] = values[row - one_day, col]
# load all data
dataset = read_csv('household_power_consumption.txt', sep=';', header=0, low_memory=False, infer_datetime_format=True, parse_dates={'datetime':[0,1]}, index_col=['datetime'])
# mark all missing values
dataset.replace('?', nan, inplace=True)
# make dataset numeric
dataset = dataset.astype('float32')
# fill missing
fill_missing(dataset.values)
# add a column for for the remainder of sub metering
values = dataset.values
dataset['sub_metering_4'] = (values[:,0] * 1000 / 60) - (values[:,4] + values[:,5] + values[:,6])
# save updated dataset
dataset.to_csv('household_power_consumption.csv')
Raw
logicbelief-master-puwork_lstm_chapter_17_02_resample_dataset.py
# resample minute data to total for each day for the power usage dataset
from pandas import read_csv
# load the new file
dataset = read_csv('household_power_consumption.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# resample data to daily
daily_groups = dataset.resample('D')
daily_data = daily_groups.sum()
# summarize
print(daily_data.shape)
print(daily_data.head())
# save
daily_data.to_csv('household_power_consumption_days.csv')
Raw
logicbelief-master-puwork_lstm_chapter_17_03_train_test_split.py
# split the power usage dataset into standard weeks
from numpy import split
from numpy import array
from pandas import read_csv
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
train, test = split_dataset(dataset.values)
# validate train data
print(train.shape)
print(train[0, 0, 0], train[-1, -1, 0])
# validate test
print(test.shape)
print(test[0, 0, 0], test[-1, -1, 0])
Raw
logicbelief-master-puwork_lstm_chapter_17_04_naive_forecasts.py
# naive forecast strategies for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# evaluate a single model
def evaluate_model(model_func, train, test):
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = model_func(history)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
predictions = array(predictions)
# evaluate predictions days for each week
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# daily persistence model
def daily_persistence(history):
# get the data for the prior week
last_week = history[-1]
# get the total active power for the last day
value = last_week[-1, 0]
# prepare 7 day forecast
forecast = [value for _ in range(7)]
return forecast
# weekly persistence model
def weekly_persistence(history):
# get the data for the prior week
last_week = history[-1]
return last_week[:, 0]
# week one year ago persistence model
def week_one_year_ago_persistence(history):
# get the data for the prior week
last_week = history[-52]
return last_week[:, 0]
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# define the names and functions for the models we wish to evaluate
models = dict()
models['daily'] = daily_persistence
models['weekly'] = weekly_persistence
models['week-oya'] = week_one_year_ago_persistence
# evaluate each model
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
for name, func in models.items():
# evaluate and get scores
score, scores = evaluate_model(func, train, test)
# summarize scores
summarize_scores(name, score, scores)
# plot scores
pyplot.plot(days, scores, marker='o', label=name)
# show plot
pyplot.legend()
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_18_01_acf_pacf_plots.py
# acf and pacf plots of total power usage
from numpy import split
from numpy import array
from pandas import read_csv
from matplotlib import pyplot
from statsmodels.graphics.tsaplots import plot_acf
from statsmodels.graphics.tsaplots import plot_pacf
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# convert windows of weekly multivariate data into a series of total power
def to_series(data):
# extract just the total power from each week
series = [week[:, 0] for week in data]
# flatten into a single series
series = array(series).flatten()
return series
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# convert training data into a series
series = to_series(train)
# plots
pyplot.figure()
lags = 365
# acf
axis = pyplot.subplot(2, 1, 1)
plot_acf(series, ax=axis, lags=lags)
# pacf
axis = pyplot.subplot(2, 1, 2)
plot_pacf(series, ax=axis, lags=lags)
# show plot
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_18_02_acf_pacf_plots_zoomed.py
# zoomed acf and pacf plots of total power usage
from numpy import split
from numpy import array
from pandas import read_csv
from matplotlib import pyplot
from statsmodels.graphics.tsaplots import plot_acf
from statsmodels.graphics.tsaplots import plot_pacf
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# convert windows of weekly multivariate data into a series of total power
def to_series(data):
# extract just the total power from each week
series = [week[:, 0] for week in data]
# flatten into a single series
series = array(series).flatten()
return series
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# convert training data into a series
series = to_series(train)
# plots
pyplot.figure()
lags = 50
# acf
axis = pyplot.subplot(2, 1, 1)
plot_acf(series, ax=axis, lags=lags)
# pacf
axis = pyplot.subplot(2, 1, 2)
plot_pacf(series, ax=axis, lags=lags)
# show plot
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_18_03_arima_model.py
# arima forecast for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from statsmodels.tsa.arima_model import ARIMA
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# evaluate a single model
def evaluate_model(model_func, train, test):
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = model_func(history)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
predictions = array(predictions)
# evaluate predictions days for each week
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# convert windows of weekly multivariate data into a series of total power
def to_series(data):
# extract just the total power from each week
series = [week[:, 0] for week in data]
# flatten into a single series
series = array(series).flatten()
return series
# arima forecast
def arima_forecast(history):
# convert history into a univariate series
series = to_series(history)
# define the model
model = ARIMA(series, order=(7,0,0))
# fit the model
model_fit = model.fit(disp=False)
# make forecast
yhat = model_fit.predict(len(series), len(series)+6)
return yhat
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# define the names and functions for the models we wish to evaluate
models = dict()
models['arima'] = arima_forecast
# evaluate each model
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
for name, func in models.items():
# evaluate and get scores
score, scores = evaluate_model(func, train, test)
# summarize scores
summarize_scores(name, score, scores)
# plot scores
pyplot.plot(days, scores, marker='o', label=name)
# show plot
pyplot.legend()
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_19_01_cnn_univariate_model.py
# univariate multi-step cnn for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
x_input = data[in_start:in_end, 0]
x_input = x_input.reshape((len(x_input), 1))
X.append(x_input)
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# train the model
def build_model(train, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 20, 4
n_timesteps, n_features, n_outputs = train_x.shape[1], train_x.shape[2], train_y.shape[1]
# define model
model = Sequential()
model.add(Conv1D(filters=16, kernel_size=3, activation='relu', input_shape=(n_timesteps,n_features)))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(10, activation='relu'))
model.add(Dense(n_outputs))
model.compile(loss='mse', optimizer='adam')
# fit network
model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, 0]
# reshape into [1, n_input, 1]
input_x = input_x.reshape((1, len(input_x), 1))
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_input):
# fit model
model = build_model(train, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# evaluate model and get scores
n_input = 7
score, scores = evaluate_model(train, test, n_input)
# summarize scores
summarize_scores('cnn', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='cnn')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_19_02_multichannel_cnn.py
# multichannel multi-step cnn for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
X.append(data[in_start:in_end, :])
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# train the model
def build_model(train, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 70, 16
n_timesteps, n_features, n_outputs = train_x.shape[1], train_x.shape[2], train_y.shape[1]
# define model
model = Sequential()
model.add(Conv1D(filters=32, kernel_size=3, activation='relu', input_shape=(n_timesteps,n_features)))
model.add(Conv1D(filters=32, kernel_size=3, activation='relu'))
model.add(MaxPooling1D(pool_size=2))
model.add(Conv1D(filters=16, kernel_size=3, activation='relu'))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs))
model.compile(loss='mse', optimizer='adam')
# fit network
model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, :]
# reshape into [1, n_input, n]
input_x = input_x.reshape((1, input_x.shape[0], input_x.shape[1]))
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_input):
# fit model
model = build_model(train, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# evaluate model and get scores
n_input = 14
score, scores = evaluate_model(train, test, n_input)
# summarize scores
summarize_scores('cnn', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='cnn')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_19_03_multiheaded_cnn.py
# multi headed multi-step cnn for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.utils.vis_utils import plot_model
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.models import Model
from keras.layers import Input
from keras.layers.merge import concatenate
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
X.append(data[in_start:in_end, :])
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# plot training history
def plot_history(history):
# plot loss
pyplot.subplot(2, 1, 1)
pyplot.plot(history.history['loss'], label='train')
pyplot.plot(history.history['val_loss'], label='test')
pyplot.title('loss', y=0, loc='center')
pyplot.legend()
# plot rmse
pyplot.subplot(2, 1, 2)
pyplot.plot(history.history['rmse'], label='train')
pyplot.plot(history.history['val_rmse'], label='test')
pyplot.title('rmse', y=0, loc='center')
pyplot.legend()
pyplot.show()
# train the model
def build_model(train, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 25, 16
n_timesteps, n_features, n_outputs = train_x.shape[1], train_x.shape[2], train_y.shape[1]
# create a channel for each variable
in_layers, out_layers = list(), list()
for _ in range(n_features):
inputs = Input(shape=(n_timesteps,1))
conv1 = Conv1D(filters=32, kernel_size=3, activation='relu')(inputs)
conv2 = Conv1D(filters=32, kernel_size=3, activation='relu')(conv1)
pool1 = MaxPooling1D(pool_size=2)(conv2)
flat = Flatten()(pool1)
# store layers
in_layers.append(inputs)
out_layers.append(flat)
# merge heads
merged = concatenate(out_layers)
# interpretation
dense1 = Dense(200, activation='relu')(merged)
dense2 = Dense(100, activation='relu')(dense1)
outputs = Dense(n_outputs)(dense2)
model = Model(inputs=in_layers, outputs=outputs)
# compile model
model.compile(loss='mse', optimizer='adam')
# plot the model
plot_model(model, show_shapes=True, to_file='multiheaded_cnn.png')
# fit network
input_data = [train_x[:,:,i].reshape((train_x.shape[0],n_timesteps,1)) for i in range(n_features)]
model.fit(input_data, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, :]
# reshape into n input arrays
input_x = [input_x[:,i].reshape((1,input_x.shape[0],1)) for i in range(input_x.shape[1])]
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_input):
# fit model
model = build_model(train, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# evaluate model and get scores
n_input = 14
score, scores = evaluate_model(train, test, n_input)
# summarize scores
summarize_scores('cnn', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='cnn')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_20_01_univariate_lstm.py
# univariate multi-step lstm for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
x_input = data[in_start:in_end, 0]
x_input = x_input.reshape((len(x_input), 1))
X.append(x_input)
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# train the model
def build_model(train, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 70, 16
n_timesteps, n_features, n_outputs = train_x.shape[1], train_x.shape[2], train_y.shape[1]
# define model
model = Sequential()
model.add(LSTM(200, activation='relu', input_shape=(n_timesteps, n_features)))
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs))
model.compile(loss='mse', optimizer='adam')
# fit network
model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, 0]
# reshape into [1, n_input, 1]
input_x = input_x.reshape((1, len(input_x), 1))
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_input):
# fit model
model = build_model(train, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# evaluate model and get scores
n_input = 7
score, scores = evaluate_model(train, test, n_input)
# summarize scores
summarize_scores('lstm', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='lstm')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_20_02_encoder_decoder_lstm.py
# univariate multi-step encoder-decoder lstm for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from keras.layers import RepeatVector
from keras.layers import TimeDistributed
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
x_input = data[in_start:in_end, 0]
x_input = x_input.reshape((len(x_input), 1))
X.append(x_input)
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# train the model
def build_model(train, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 20, 16
n_timesteps, n_features, n_outputs = train_x.shape[1], train_x.shape[2], train_y.shape[1]
# reshape output into [samples, timesteps, features]
train_y = train_y.reshape((train_y.shape[0], train_y.shape[1], 1))
# define model
model = Sequential()
model.add(LSTM(200, activation='relu', input_shape=(n_timesteps, n_features)))
model.add(RepeatVector(n_outputs))
model.add(LSTM(200, activation='relu', return_sequences=True))
model.add(TimeDistributed(Dense(100, activation='relu')))
model.add(TimeDistributed(Dense(1)))
model.compile(loss='mse', optimizer='adam')
# fit network
model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, 0]
# reshape into [1, n_input, 1]
input_x = input_x.reshape((1, len(input_x), 1))
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_input):
# fit model
model = build_model(train, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# evaluate model and get scores
n_input = 14
score, scores = evaluate_model(train, test, n_input)
# summarize scores
summarize_scores('lstm', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='lstm')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_20_03_multivariate_encoder_decoder_lstm.py
# multivariate multi-step encoder-decoder lstm for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from keras.layers import RepeatVector
from keras.layers import TimeDistributed
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
X.append(data[in_start:in_end, :])
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# train the model
def build_model(train, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 50, 16
n_timesteps, n_features, n_outputs = train_x.shape[1], train_x.shape[2], train_y.shape[1]
# reshape output into [samples, timesteps, features]
train_y = train_y.reshape((train_y.shape[0], train_y.shape[1], 1))
# define model
model = Sequential()
model.add(LSTM(200, activation='relu', input_shape=(n_timesteps, n_features)))
model.add(RepeatVector(n_outputs))
model.add(LSTM(200, activation='relu', return_sequences=True))
model.add(TimeDistributed(Dense(100, activation='relu')))
model.add(TimeDistributed(Dense(1)))
model.compile(loss='mse', optimizer='adam')
# fit network
model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, :]
# reshape into [1, n_input, n]
input_x = input_x.reshape((1, input_x.shape[0], input_x.shape[1]))
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_input):
# fit model
model = build_model(train, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# evaluate model and get scores
n_input = 14
score, scores = evaluate_model(train, test, n_input)
# summarize scores
summarize_scores('lstm', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='lstm')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_20_04_cnn_encoder_decoder_lstm.py
# univariate multi-step encoder-decoder cnn-lstm for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import LSTM
from keras.layers import RepeatVector
from keras.layers import TimeDistributed
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
x_input = data[in_start:in_end, 0]
x_input = x_input.reshape((len(x_input), 1))
X.append(x_input)
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# train the model
def build_model(train, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 20, 16
n_timesteps, n_features, n_outputs = train_x.shape[1], train_x.shape[2], train_y.shape[1]
# reshape output into [samples, timesteps, features]
train_y = train_y.reshape((train_y.shape[0], train_y.shape[1], 1))
# define model
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=3, activation='relu', input_shape=(n_timesteps,n_features)))
model.add(Conv1D(filters=64, kernel_size=3, activation='relu'))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(RepeatVector(n_outputs))
model.add(LSTM(200, activation='relu', return_sequences=True))
model.add(TimeDistributed(Dense(100, activation='relu')))
model.add(TimeDistributed(Dense(1)))
model.compile(loss='mse', optimizer='adam')
# fit network
model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, 0]
# reshape into [1, n_input, 1]
input_x = input_x.reshape((1, len(input_x), 1))
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_input):
# fit model
model = build_model(train, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# evaluate model and get scores
n_input = 14
score, scores = evaluate_model(train, test, n_input)
# summarize scores
summarize_scores('lstm', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='lstm')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_20_05_convlstm_encoder_decoder_lstm.py
# univariate multi-step encoder-decoder convlstm for the power usage dataset
from math import sqrt
from numpy import split
from numpy import array
from pandas import read_csv
from sklearn.metrics import mean_squared_error
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import LSTM
from keras.layers import RepeatVector
from keras.layers import TimeDistributed
from keras.layers import ConvLSTM2D
# split a univariate dataset into train/test sets
def split_dataset(data):
# split into standard weeks
train, test = data[1:-328], data[-328:-6]
# restructure into windows of weekly data
train = array(split(train, len(train)/7))
test = array(split(test, len(test)/7))
return train, test
# evaluate one or more weekly forecasts against expected values
def evaluate_forecasts(actual, predicted):
scores = list()
# calculate an RMSE score for each day
for i in range(actual.shape[1]):
# calculate mse
mse = mean_squared_error(actual[:, i], predicted[:, i])
# calculate rmse
rmse = sqrt(mse)
# store
scores.append(rmse)
# calculate overall RMSE
s = 0
for row in range(actual.shape[0]):
for col in range(actual.shape[1]):
s += (actual[row, col] - predicted[row, col])**2
score = sqrt(s / (actual.shape[0] * actual.shape[1]))
return score, scores
# summarize scores
def summarize_scores(name, score, scores):
s_scores = ', '.join(['%.1f' % s for s in scores])
print('%s: [%.3f] %s' % (name, score, s_scores))
# convert history into inputs and outputs
def to_supervised(train, n_input, n_out=7):
# flatten data
data = train.reshape((train.shape[0]*train.shape[1], train.shape[2]))
X, y = list(), list()
in_start = 0
# step over the entire history one time step at a time
for _ in range(len(data)):
# define the end of the input sequence
in_end = in_start + n_input
out_end = in_end + n_out
# ensure we have enough data for this instance
if out_end < len(data):
x_input = data[in_start:in_end, 0]
x_input = x_input.reshape((len(x_input), 1))
X.append(x_input)
y.append(data[in_end:out_end, 0])
# move along one time step
in_start += 1
return array(X), array(y)
# train the model
def build_model(train, n_steps, n_length, n_input):
# prepare data
train_x, train_y = to_supervised(train, n_input)
# define parameters
verbose, epochs, batch_size = 0, 20, 16
n_features, n_outputs = train_x.shape[2], train_y.shape[1]
# reshape into subsequences [samples, timesteps, rows, cols, channels]
train_x = train_x.reshape((train_x.shape[0], n_steps, 1, n_length, n_features))
# reshape output into [samples, timesteps, features]
train_y = train_y.reshape((train_y.shape[0], train_y.shape[1], 1))
# define model
model = Sequential()
model.add(ConvLSTM2D(filters=64, kernel_size=(1,3), activation='relu', input_shape=(n_steps, 1, n_length, n_features)))
model.add(Flatten())
model.add(RepeatVector(n_outputs))
model.add(LSTM(200, activation='relu', return_sequences=True))
model.add(TimeDistributed(Dense(100, activation='relu')))
model.add(TimeDistributed(Dense(1)))
model.compile(loss='mse', optimizer='adam')
# fit network
model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, verbose=verbose)
return model
# make a forecast
def forecast(model, history, n_steps, n_length, n_input):
# flatten data
data = array(history)
data = data.reshape((data.shape[0]*data.shape[1], data.shape[2]))
# retrieve last observations for input data
input_x = data[-n_input:, 0]
# reshape into [samples, timesteps, rows, cols, channels]
input_x = input_x.reshape((1, n_steps, 1, n_length, 1))
# forecast the next week
yhat = model.predict(input_x, verbose=0)
# we only want the vector forecast
yhat = yhat[0]
return yhat
# evaluate a single model
def evaluate_model(train, test, n_steps, n_length, n_input):
# fit model
model = build_model(train, n_steps, n_length, n_input)
# history is a list of weekly data
history = [x for x in train]
# walk-forward validation over each week
predictions = list()
for i in range(len(test)):
# predict the week
yhat_sequence = forecast(model, history, n_steps, n_length, n_input)
# store the predictions
predictions.append(yhat_sequence)
# get real observation and add to history for predicting the next week
history.append(test[i, :])
# evaluate predictions days for each week
predictions = array(predictions)
score, scores = evaluate_forecasts(test[:, :, 0], predictions)
return score, scores
# load the new file
dataset = read_csv('household_power_consumption_days.csv', header=0, infer_datetime_format=True, parse_dates=['datetime'], index_col=['datetime'])
# split into train and test
train, test = split_dataset(dataset.values)
# define the number of subsequences and the length of subsequences
n_steps, n_length = 2, 7
# define the total days to use as input
n_input = n_length * n_steps
score, scores = evaluate_model(train, test, n_steps, n_length, n_input)
# summarize scores
summarize_scores('lstm', score, scores)
# plot scores
days = ['sun', 'mon', 'tue', 'wed', 'thr', 'fri', 'sat']
pyplot.plot(days, scores, marker='o', label='lstm')
pyplot.show()
Raw
logicbelief-master-puwork_lstm_chapter_22_01_load_file.py
# load one file from the har dataset
from pandas import read_csv
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
data = load_file('HARDataset/train/Inertial Signals/total_acc_y_train.txt')
print(data.shape)
Raw
logicbelief-master-puwork_lstm_chapter_22_02_load_group_of_files.py
# load group of files from the har dataset
from numpy import dstack
from pandas import read_csv
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files, such as x, y, z data for a given variable
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load the total acc data
filenames = ['total_acc_x_train.txt', 'total_acc_y_train.txt', 'total_acc_z_train.txt']
total_acc = load_group(filenames, prefix='HARDataset/train/Inertial Signals/')
print(total_acc.shape)
Raw
logicbelief-master-puwork_lstm_chapter_22_03_load_all_files.py
# load all train and test data from the har dataset
from numpy import dstack
from pandas import read_csv
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files, such as x, y, z data for a given variable
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load all train
trainX, trainy = load_dataset('train', 'HARDataset/')
print(trainX.shape, trainy.shape)
# load all test
testX, testy = load_dataset('test', 'HARDataset/')
print(testX.shape, testy.shape)
Raw
logicbelief-master-puwork_lstm_chapter_22_04_class_breakdown.py
# summarize class balance from the har dataset
from numpy import vstack
from pandas import read_csv
from pandas import DataFrame
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# summarize the balance of classes in an output variable column
def class_breakdown(data):
# convert the numpy array into a dataframe
df = DataFrame(data)
# group data by the class value and calculate the number of rows
counts = df.groupby(0).size()
# retrieve raw rows
counts = counts.values
# summarize
for i in range(len(counts)):
percent = counts[i] / len(df) * 100
print('Class=%d, total=%d, percentage=%.3f' % (i+1, counts[i], percent))
# load train file
trainy = load_file('HARDataset/train/y_train.txt')
# summarize class breakdown
print('Train Dataset')
class_breakdown(trainy)
# load test file
testy = load_file('HARDataset/test/y_test.txt')
# summarize class breakdown
print('Test Dataset')
class_breakdown(testy)
# summarize combined class breakdown
print('Both')
combined = vstack((trainy, testy))
class_breakdown(combined)
Raw
logicbelief-master-puwork_lstm_chapter_22_05_plot_data_for_subject.py
# plot all vars for one subject in the har dataset
from numpy import dstack
from numpy import unique
from pandas import read_csv
from matplotlib import pyplot
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files, such as x, y, z data for a given variable
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# get all data for one subject
def data_for_subject(X, y, sub_map, sub_id):
# get row indexes for the subject id
ix = [i for i in range(len(sub_map)) if sub_map[i]==sub_id]
# return the selected samples
return X[ix, :, :], y[ix]
# convert a series of windows to a 1D list
def to_series(windows):
series = list()
for window in windows:
# remove the overlap from the window
half = int(len(window) / 2) - 1
for value in window[-half:]:
series.append(value)
return series
# plot the data for one subject
def plot_subject(X, y):
pyplot.figure()
# determine the total number of plots
n, off = X.shape[2] + 1, 0
# plot total acc
for i in range(3):
pyplot.subplot(n, 1, off+1)
pyplot.plot(to_series(X[:, :, off]))
pyplot.title('total acc '+str(i), y=0, loc='left', size=7)
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
off += 1
# plot body acc
for i in range(3):
pyplot.subplot(n, 1, off+1)
pyplot.plot(to_series(X[:, :, off]))
pyplot.title('body acc '+str(i), y=0, loc='left', size=7)
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
off += 1
# plot body gyro
for i in range(3):
pyplot.subplot(n, 1, off+1)
pyplot.plot(to_series(X[:, :, off]))
pyplot.title('body gyro '+str(i), y=0, loc='left', size=7)
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
off += 1
# plot activities
pyplot.subplot(n, 1, n)
pyplot.plot(y)
pyplot.title('activity', y=0, loc='left', size=7)
# turn off ticks to remove clutter
pyplot.yticks([])
pyplot.xticks([])
pyplot.show()
# load data
trainX, trainy = load_dataset('train', 'HARDataset/')
# load mapping of rows to subjects
sub_map = load_file('HARDataset/train/subject_train.txt')
train_subjects = unique(sub_map)
print(train_subjects)
# get the data for one subject
sub_id = train_subjects[0]
subX, suby = data_for_subject(trainX, trainy, sub_map, sub_id)
print(subX.shape, suby.shape)
# plot data for subject
plot_subject(subX, suby)
Raw
logicbelief-master-puwork_lstm_chapter_22_06_plot_histograms_for_subjects.py
# plot histograms for multiple subjects from the har dataset
from numpy import unique
from numpy import dstack
from pandas import read_csv
from matplotlib import pyplot
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files, such as x, y, z data for a given variable
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# get all data for one subject
def data_for_subject(X, y, sub_map, sub_id):
# get row indexes for the subject id
ix = [i for i in range(len(sub_map)) if sub_map[i]==sub_id]
# return the selected samples
return X[ix, :, :], y[ix]
# convert a series of windows to a 1D list
def to_series(windows):
series = list()
for window in windows:
# remove the overlap from the window
half = int(len(window) / 2) - 1
for value in window[-half:]:
series.append(value)
return series
# plot histograms for multiple subjects
def plot_subject_histograms(X, y, sub_map, offset, n=10):
pyplot.figure()
# get unique subjects
subject_ids = unique(sub_map[:,0])
# enumerate subjects
for k in range(n):
sub_id = subject_ids[k]
# get data for one subject
subX, _ = data_for_subject(X, y, sub_map, sub_id)
# total acc
for i in range(3):
ax = pyplot.subplot(n, 1, k+1)
ax.set_xlim(-1,1)
ax.hist(to_series(subX[:,:,offset+i]), bins=100)
pyplot.yticks([])
pyplot.xticks([-1,0,1])
pyplot.show()
# load training dataset
X, y = load_dataset('train', 'HARDataset/')
# load mapping of rows to subjects
sub_map = load_file('HARDataset/train/subject_train.txt')
# plot total acceleration histograms for subjects
plot_subject_histograms(X, y, sub_map, 0)
# plot body acceleration histograms for subjects
plot_subject_histograms(X, y, sub_map, 3)
# plot gyroscopic histograms for subjects
plot_subject_histograms(X, y, sub_map, 6)
Raw
logicbelief-master-puwork_lstm_chapter_22_07_plot_histograms_by_activity.py
# plot histograms per activity for a subject from the har dataset
from numpy import dstack
from numpy import unique
from pandas import read_csv
from matplotlib import pyplot
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files, such as x, y, z data for a given variable
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# get all data for one subject
def data_for_subject(X, y, sub_map, sub_id):
# get row indexes for the subject id
ix = [i for i in range(len(sub_map)) if sub_map[i]==sub_id]
# return the selected samples
return X[ix, :, :], y[ix]
# convert a series of windows to a 1D list
def to_series(windows):
series = list()
for window in windows:
# remove the overlap from the window
half = int(len(window) / 2) - 1
for value in window[-half:]:
series.append(value)
return series
# group data by activity
def data_by_activity(X, y, activities):
# group windows by activity
return {a:X[y[:,0]==a, :, :] for a in activities}
# plot histograms for each activity for a subject
def plot_activity_histograms(X, y, offset):
# get a list of unique activities for the subject
activity_ids = unique(y[:,0])
# group windows by activity
grouped = data_by_activity(X, y, activity_ids)
# plot per activity, histograms for each axis
pyplot.figure()
for k in range(len(activity_ids)):
act_id = activity_ids[k]
# total acceleration
for i in range(3):
ax = pyplot.subplot(len(activity_ids), 1, k+1)
ax.set_xlim(-1,1)
# create histogra,
pyplot.hist(to_series(grouped[act_id][:,:,offset+i]), bins=100)
# create title
pyplot.title('activity '+str(act_id), y=0, loc='left', size=10)
# simplify axis
pyplot.yticks([])
pyplot.xticks([-1,0,1])
pyplot.show()
# load data
trainX, trainy = load_dataset('train', 'HARDataset/')
# load mapping of rows to subjects
sub_map = load_file('HARDataset/train/subject_train.txt')
train_subjects = unique(sub_map)
# get the data for one subject
sub_id = train_subjects[0]
subX, suby = data_for_subject(trainX, trainy, sub_map, sub_id)
# plot total acceleration histograms per activity for a subject
plot_activity_histograms(subX, suby, 0)
# plot body acceleration histograms per activity for a subject
plot_activity_histograms(subX, suby, 3)
# plot gyroscopic histograms per activity for a subject
plot_activity_histograms(subX, suby, 6)
Raw
logicbelief-master-puwork_lstm_chapter_22_08_plot_activity_durations.py
# plot durations of each activity by subject from the har dataset
from numpy import dstack
from numpy import unique
from pandas import read_csv
from matplotlib import pyplot
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files, such as x, y, z data for a given variable
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# get all data for one subject
def data_for_subject(X, y, sub_map, sub_id):
# get row indexes for the subject id
ix = [i for i in range(len(sub_map)) if sub_map[i]==sub_id]
# return the selected samples
return X[ix, :, :], y[ix]
# convert a series of windows to a 1D list
def to_series(windows):
series = list()
for window in windows:
# remove the overlap from the window
half = int(len(window) / 2) - 1
for value in window[-half:]:
series.append(value)
return series
# group data by activity
def data_by_activity(X, y, activities):
# group windows by activity
return {a:X[y[:,0]==a, :, :] for a in activities}
# plot activity durations by subject
def plot_activity_durations_by_subject(X, y, sub_map):
# get unique subjects and activities
subject_ids = unique(sub_map[:,0])
activity_ids = unique(y[:,0])
# enumerate subjects
activity_windows = {a:list() for a in activity_ids}
for sub_id in subject_ids:
# get data for one subject
_, subj_y = data_for_subject(X, y, sub_map, sub_id)
# count windows by activity
for a in activity_ids:
activity_windows[a].append(len(subj_y[subj_y[:,0]==a]))
# organize durations into a list of lists
durations = [activity_windows[a] for a in activity_ids]
pyplot.boxplot(durations, labels=activity_ids)
pyplot.show()
# load training dataset
X, y = load_dataset('train', 'HARDataset/')
# load mapping of rows to subjects
sub_map = load_file('HARDataset/train/subject_train.txt')
# plot durations
plot_activity_durations_by_subject(X, y, sub_map)
Raw
logicbelief-master-puwork_lstm_chapter_23_01_evaluate_ml_models.py
# spot check ml algorithms on engineered-features from the har dataset
from pandas import read_csv
from sklearn.metrics import accuracy_score
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.svm import SVC
from sklearn.naive_bayes import GaussianNB
from sklearn.ensemble import BaggingClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import ExtraTreesClassifier
from sklearn.ensemble import GradientBoostingClassifier
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
# load input data
X = load_file(prefix + group + '/X_'+group+'.txt')
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# flatten y
trainy, testy = trainy[:,0], testy[:,0]
return trainX, trainy, testX, testy
# create a dict of standard models to evaluate {name:object}
def define_models(models=dict()):
# nonlinear models
models['knn'] = KNeighborsClassifier(n_neighbors=7)
models['cart'] = DecisionTreeClassifier()
models['svm'] = SVC()
models['bayes'] = GaussianNB()
# ensemble models
models['bag'] = BaggingClassifier(n_estimators=100)
models['rf'] = RandomForestClassifier(n_estimators=100)
models['et'] = ExtraTreesClassifier(n_estimators=100)
models['gbm'] = GradientBoostingClassifier(n_estimators=100)
print('Defined %d models' % len(models))
return models
# evaluate a single model
def evaluate_model(trainX, trainy, testX, testy, model):
# fit the model
model.fit(trainX, trainy)
# make predictions
yhat = model.predict(testX)
# evaluate predictions
accuracy = accuracy_score(testy, yhat)
return accuracy * 100.0
# evaluate a dict of models {name:object}, returns {name:score}
def evaluate_models(trainX, trainy, testX, testy, models):
results = dict()
for name, model in models.items():
# evaluate the model
results[name] = evaluate_model(trainX, trainy, testX, testy, model)
# show process
print('>%s: %.3f' % (name, results[name]))
return results
# print and plot the results
def summarize_results(results, maximize=True):
# create a list of (name, mean(scores)) tuples
mean_scores = [(k,v) for k,v in results.items()]
# sort tuples by mean score
mean_scores = sorted(mean_scores, key=lambda x: x[1])
# reverse for descending order (e.g. for accuracy)
if maximize:
mean_scores = list(reversed(mean_scores))
print()
for name, score in mean_scores:
print('Name=%s, Score=%.3f' % (name, score))
# load dataset
trainX, trainy, testX, testy = load_dataset()
# get model list
models = define_models()
# evaluate models
results = evaluate_models(trainX, trainy, testX, testy, models)
# summarize results
summarize_results(results)
Raw
logicbelief-master-puwork_lstm_chapter_23_02_evaluate_ml_models_raw_data.py
# spot check on raw data from the har dataset
from numpy import dstack
from pandas import read_csv
from sklearn.metrics import accuracy_score
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.svm import SVC
from sklearn.naive_bayes import GaussianNB
from sklearn.ensemble import BaggingClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import ExtraTreesClassifier
from sklearn.ensemble import GradientBoostingClassifier
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files into a 3D array of [samples, timesteps, features]
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# flatten X
trainX = trainX.reshape((trainX.shape[0], trainX.shape[1] * trainX.shape[2]))
testX = testX.reshape((testX.shape[0], testX.shape[1] * testX.shape[2]))
# flatten y
trainy, testy = trainy[:,0], testy[:,0]
return trainX, trainy, testX, testy
# create a dict of standard models to evaluate {name:object}
def define_models(models=dict()):
# nonlinear models
models['knn'] = KNeighborsClassifier(n_neighbors=7)
models['cart'] = DecisionTreeClassifier()
models['svm'] = SVC()
models['bayes'] = GaussianNB()
# ensemble models
models['bag'] = BaggingClassifier(n_estimators=100)
models['rf'] = RandomForestClassifier(n_estimators=100)
models['et'] = ExtraTreesClassifier(n_estimators=100)
models['gbm'] = GradientBoostingClassifier(n_estimators=100)
print('Defined %d models' % len(models))
return models
# evaluate a single model
def evaluate_model(trainX, trainy, testX, testy, model):
# fit the model
model.fit(trainX, trainy)
# make predictions
yhat = model.predict(testX)
# evaluate predictions
accuracy = accuracy_score(testy, yhat)
return accuracy * 100.0
# evaluate a dict of models {name:object}, returns {name:score}
def evaluate_models(trainX, trainy, testX, testy, models):
results = dict()
for name, model in models.items():
# evaluate the model
results[name] = evaluate_model(trainX, trainy, testX, testy, model)
# show process
print('>%s: %.3f' % (name, results[name]))
return results
# print and plot the results
def summarize_results(results, maximize=True):
# create a list of (name, mean(scores)) tuples
mean_scores = [(k,v) for k,v in results.items()]
# sort tuples by mean score
mean_scores = sorted(mean_scores, key=lambda x: x[1])
# reverse for descending order (e.g. for accuracy)
if maximize:
mean_scores = list(reversed(mean_scores))
print()
for name, score in mean_scores:
print('Name=%s, Score=%.3f' % (name, score))
# load dataset
trainX, trainy, testX, testy = load_dataset()
# get model list
models = define_models()
# evaluate models
results = evaluate_models(trainX, trainy, testX, testy, models)
# summarize results
summarize_results(results)
Raw
logicbelief-master-puwork_lstm_chapter_24_01_1d_cnn_model.py
# cnn model for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import Dropout
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.utils import to_categorical
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy):
verbose, epochs, batch_size = 0, 10, 32
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[2], trainy.shape[1]
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=3, activation='relu', input_shape=(n_timesteps,n_features)))
model.add(Conv1D(filters=64, kernel_size=3, activation='relu'))
model.add(Dropout(0.5))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores):
print(scores)
m, s = mean(scores), std(scores)
print('Accuracy: %.3f%% (+/-%.3f)' % (m, s))
# run an experiment
def run_experiment(repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy)
score = score * 100.0
print('>#%d: %.3f' % (r+1, score))
scores.append(score)
# summarize results
summarize_results(scores)
# run the experiment
run_experiment()
Raw
logicbelief-master-puwork_lstm_chapter_24_02_histograms_all_variables.py
# plot distributions for the har dataset
from numpy import dstack
from pandas import read_csv
from keras.utils import to_categorical
from matplotlib import pyplot
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# plot a histogram of each variable in the dataset
def plot_variable_distributions(trainX):
# remove overlap
cut = int(trainX.shape[1] / 2)
longX = trainX[:, -cut:, :]
# flatten windows
longX = longX.reshape((longX.shape[0] * longX.shape[1], longX.shape[2]))
pyplot.figure()
for i in range(longX.shape[1]):
# create figure
ax = pyplot.subplot(longX.shape[1], 1, i+1)
ax.set_xlim(-1, 1)
# create histogram
pyplot.hist(longX[:, i], bins=100)
# simplify axis remove clutter
pyplot.yticks([])
pyplot.xticks([-1,0,1])
pyplot.show()
# load data
trainX, trainy, testX, testy = load_dataset()
# plot histograms
plot_variable_distributions(trainX)
Raw
logicbelief-master-puwork_lstm_chapter_24_03_cnn_standardization.py
# cnn model with standardization for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from matplotlib import pyplot
from sklearn.preprocessing import StandardScaler
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import Dropout
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.utils import to_categorical
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# standardize data
def scale_data(trainX, testX, standardize):
# remove overlap
cut = int(trainX.shape[1] / 2)
longX = trainX[:, -cut:, :]
# flatten windows
longX = longX.reshape((longX.shape[0] * longX.shape[1], longX.shape[2]))
# flatten train and test
flatTrainX = trainX.reshape((trainX.shape[0] * trainX.shape[1], trainX.shape[2]))
flatTestX = testX.reshape((testX.shape[0] * testX.shape[1], testX.shape[2]))
# standardize
if standardize:
s = StandardScaler()
# fit on training data
s.fit(longX)
# apply to training and test data
longX = s.transform(longX)
flatTrainX = s.transform(flatTrainX)
flatTestX = s.transform(flatTestX)
# reshape
flatTrainX = flatTrainX.reshape((trainX.shape))
flatTestX = flatTestX.reshape((testX.shape))
return flatTrainX, flatTestX
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy, param):
verbose, epochs, batch_size = 0, 10, 32
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[2], trainy.shape[1]
# scale data
trainX, testX = scale_data(trainX, testX, param)
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=3, activation='relu', input_shape=(n_timesteps,n_features)))
model.add(Conv1D(filters=64, kernel_size=3, activation='relu'))
model.add(Dropout(0.5))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores, params):
print(scores, params)
# summarize mean and standard deviation
for i in range(len(scores)):
m, s = mean(scores[i]), std(scores[i])
print('Param=%s: %.3f%% (+/-%.3f)' % (params[i], m, s))
# boxplot of scores
pyplot.boxplot(scores, labels=params)
pyplot.savefig('exp_cnn_standardize.png')
# run an experiment
def run_experiment(params, repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# test each parameter
all_scores = list()
for p in params:
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy, p)
score = score * 100.0
print('>p=%s #%d: %.3f' % (p, r+1, score))
scores.append(score)
all_scores.append(scores)
# summarize results
summarize_results(all_scores, params)
# run the experiment
n_params = [False, True]
run_experiment(n_params)
Raw
logicbelief-master-puwork_lstm_chapter_24_04_cnn_tune_filter_maps.py
# cnn model with filters for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import Dropout
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.utils import to_categorical
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy, n_filters):
verbose, epochs, batch_size = 0, 10, 32
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[2], trainy.shape[1]
model = Sequential()
model.add(Conv1D(filters=n_filters, kernel_size=3, activation='relu', input_shape=(n_timesteps,n_features)))
model.add(Conv1D(filters=n_filters, kernel_size=3, activation='relu'))
model.add(Dropout(0.5))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores, params):
print(scores, params)
# summarize mean and standard deviation
for i in range(len(scores)):
m, s = mean(scores[i]), std(scores[i])
print('Param=%d: %.3f%% (+/-%.3f)' % (params[i], m, s))
# boxplot of scores
pyplot.boxplot(scores, labels=params)
pyplot.savefig('exp_cnn_filters.png')
# run an experiment
def run_experiment(params, repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# test each parameter
all_scores = list()
for p in params:
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy, p)
score = score * 100.0
print('>p=%d #%d: %.3f' % (p, r+1, score))
scores.append(score)
all_scores.append(scores)
# summarize results
summarize_results(all_scores, params)
# run the experiment
n_params = [8, 16, 32, 64, 128, 256]
run_experiment(n_params)
Raw
logicbelief-master-puwork_lstm_chapter_24_05_cnn_tune_kernel_size.py
# cnn model vary kernel size for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from matplotlib import pyplot
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import Dropout
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.utils import to_categorical
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy, n_kernel):
verbose, epochs, batch_size = 0, 15, 32
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[2], trainy.shape[1]
model = Sequential()
model.add(Conv1D(filters=64, kernel_size=n_kernel, activation='relu', input_shape=(n_timesteps,n_features)))
model.add(Conv1D(filters=64, kernel_size=n_kernel, activation='relu'))
model.add(Dropout(0.5))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores, params):
print(scores, params)
# summarize mean and standard deviation
for i in range(len(scores)):
m, s = mean(scores[i]), std(scores[i])
print('Param=%d: %.3f%% (+/-%.3f)' % (params[i], m, s))
# boxplot of scores
pyplot.boxplot(scores, labels=params)
pyplot.savefig('exp_cnn_kernel.png')
# run an experiment
def run_experiment(params, repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# test each parameter
all_scores = list()
for p in params:
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy, p)
score = score * 100.0
print('>p=%d #%d: %.3f' % (p, r+1, score))
scores.append(score)
all_scores.append(scores)
# summarize results
summarize_results(all_scores, params)
# run the experiment
n_params = [2, 3, 5, 7, 11]
run_experiment(n_params)
Raw
logicbelief-master-puwork_lstm_chapter_24_06_multiheaded_cnn.py
# multi-headed cnn model for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from keras.utils import to_categorical
from keras.utils.vis_utils import plot_model
from keras.models import Model
from keras.layers import Input
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import Dropout
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.layers.merge import concatenate
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy):
verbose, epochs, batch_size = 0, 10, 32
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[2], trainy.shape[1]
# head 1
inputs1 = Input(shape=(n_timesteps,n_features))
conv1 = Conv1D(filters=64, kernel_size=3, activation='relu')(inputs1)
drop1 = Dropout(0.5)(conv1)
pool1 = MaxPooling1D(pool_size=2)(drop1)
flat1 = Flatten()(pool1)
# head 2
inputs2 = Input(shape=(n_timesteps,n_features))
conv2 = Conv1D(filters=64, kernel_size=5, activation='relu')(inputs2)
drop2 = Dropout(0.5)(conv2)
pool2 = MaxPooling1D(pool_size=2)(drop2)
flat2 = Flatten()(pool2)
# head 3
inputs3 = Input(shape=(n_timesteps,n_features))
conv3 = Conv1D(filters=64, kernel_size=11, activation='relu')(inputs3)
drop3 = Dropout(0.5)(conv3)
pool3 = MaxPooling1D(pool_size=2)(drop3)
flat3 = Flatten()(pool3)
# merge
merged = concatenate([flat1, flat2, flat3])
# interpretation
dense1 = Dense(100, activation='relu')(merged)
outputs = Dense(n_outputs, activation='softmax')(dense1)
model = Model(inputs=[inputs1, inputs2, inputs3], outputs=outputs)
# save a plot of the model
plot_model(model, show_shapes=True, to_file='multiheaded.png')
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit([trainX,trainX,trainX], trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate([testX,testX,testX], testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores):
print(scores)
m, s = mean(scores), std(scores)
print('Accuracy: %.3f%% (+/-%.3f)' % (m, s))
# run an experiment
def run_experiment(repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy)
score = score * 100.0
print('>#%d: %.3f' % (r+1, score))
scores.append(score)
# summarize results
summarize_results(scores)
# run the experiment
run_experiment()
Raw
logicbelief-master-puwork_lstm_chapter_25_01_lstm_model.py
# lstm model for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Dropout
from keras.layers import LSTM
from keras.utils import to_categorical
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy):
verbose, epochs, batch_size = 0, 15, 64
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[2], trainy.shape[1]
model = Sequential()
model.add(LSTM(100, input_shape=(n_timesteps,n_features)))
model.add(Dropout(0.5))
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores):
print(scores)
m, s = mean(scores), std(scores)
print('Accuracy: %.3f%% (+/-%.3f)' % (m, s))
# run an experiment
def run_experiment(repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy)
score = score * 100.0
print('>#%d: %.3f' % (r+1, score))
scores.append(score)
# summarize results
summarize_results(scores)
# run the experiment
run_experiment()
Raw
logicbelief-master-puwork_lstm_chapter_25_02_cnn_lstm_model.py
# cnn lstm model for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import Dropout
from keras.layers import LSTM
from keras.layers import TimeDistributed
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.utils import to_categorical
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy):
# define model
verbose, epochs, batch_size = 0, 25, 64
n_features, n_outputs = trainX.shape[2], trainy.shape[1]
# reshape data into time steps of sub-sequences
n_steps, n_length = 4, 32
trainX = trainX.reshape((trainX.shape[0], n_steps, n_length, n_features))
testX = testX.reshape((testX.shape[0], n_steps, n_length, n_features))
# define model
model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=3, activation='relu'), input_shape=(None,n_length,n_features)))
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=3, activation='relu')))
model.add(TimeDistributed(Dropout(0.5)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dropout(0.5))
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores):
print(scores)
m, s = mean(scores), std(scores)
print('Accuracy: %.3f%% (+/-%.3f)' % (m, s))
# run an experiment
def run_experiment(repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy)
score = score * 100.0
print('>#%d: %.3f' % (r+1, score))
scores.append(score)
# summarize results
summarize_results(scores)
# run the experiment
run_experiment()
Raw
logicbelief-master-puwork_lstm_chapter_25_03_convlstm_model.py
# convlstm model for the har dataset
from numpy import mean
from numpy import std
from numpy import dstack
from pandas import read_csv
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
from keras.layers import Dropout
from keras.layers import ConvLSTM2D
from keras.utils import to_categorical
# load a single file as a numpy array
def load_file(filepath):
dataframe = read_csv(filepath, header=None, delim_whitespace=True)
return dataframe.values
# load a list of files and return as a 3d numpy array
def load_group(filenames, prefix=''):
loaded = list()
for name in filenames:
data = load_file(prefix + name)
loaded.append(data)
# stack group so that features are the 3rd dimension
loaded = dstack(loaded)
return loaded
# load a dataset group, such as train or test
def load_dataset_group(group, prefix=''):
filepath = prefix + group + '/Inertial Signals/'
# load all 9 files as a single array
filenames = list()
# total acceleration
filenames += ['total_acc_x_'+group+'.txt', 'total_acc_y_'+group+'.txt', 'total_acc_z_'+group+'.txt']
# body acceleration
filenames += ['body_acc_x_'+group+'.txt', 'body_acc_y_'+group+'.txt', 'body_acc_z_'+group+'.txt']
# body gyroscope
filenames += ['body_gyro_x_'+group+'.txt', 'body_gyro_y_'+group+'.txt', 'body_gyro_z_'+group+'.txt']
# load input data
X = load_group(filenames, filepath)
# load class output
y = load_file(prefix + group + '/y_'+group+'.txt')
return X, y
# load the dataset, returns train and test X and y elements
def load_dataset(prefix=''):
# load all train
trainX, trainy = load_dataset_group('train', prefix + 'HARDataset/')
# load all test
testX, testy = load_dataset_group('test', prefix + 'HARDataset/')
# zero-offset class values
trainy = trainy - 1
testy = testy - 1
# one hot encode y
trainy = to_categorical(trainy)
testy = to_categorical(testy)
return trainX, trainy, testX, testy
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy):
# define model
verbose, epochs, batch_size = 0, 25, 64
n_features, n_outputs = trainX.shape[2], trainy.shape[1]
# reshape into subsequences (samples, time steps, rows, cols, channels)
n_steps, n_length = 4, 32
trainX = trainX.reshape((trainX.shape[0], n_steps, 1, n_length, n_features))
testX = testX.reshape((testX.shape[0], n_steps, 1, n_length, n_features))
# define model
model = Sequential()
model.add(ConvLSTM2D(filters=64, kernel_size=(1,3), activation='relu', input_shape=(n_steps, 1, n_length, n_features)))
model.add(Dropout(0.5))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores):
print(scores)
m, s = mean(scores), std(scores)
print('Accuracy: %.3f%% (+/-%.3f)' % (m, s))
# run an experiment
def run_experiment(repeats=10):
# load data
trainX, trainy, testX, testy = load_dataset()
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy)
score = score * 100.0
print('>#%d: %.3f' % (r+1, score))
scores.append(score)
# summarize results
summarize_results(scores)
# run the experiment
run_experiment()
Raw
logicbelief-master-puwork_lstm_lb_stock.py
# load and plot daily births dataset
from numpy import array
from numpy import mean
from numpy import std
from pandas import read_csv
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from keras.utils.np_utils import to_categorical
import warnings
warnings.simplefilter(action='ignore', category=FutureWarning)
warnings.simplefilter(action='ignore', category=DeprecationWarning)
def split_sequences(sequences, n_steps):
X = list()
for i in range(len(sequences)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the dataset
if end_ix > len(sequences) - 1:
break
# gather input and output parts of the pattern
seq_x = sequences[i:end_ix, :]
X.append(seq_x)
return array(X)
# load
def load_dataset(nday, n_steps):
series = read_csv('/Users/ruoang/Company/astocks/600868.SH.csv', header=0, index_col=0) # summarize shape
nday_close = series['Close'].shift(nday)
nday_pct_chg = (nday_close - series['Close']) > 0
nday_pct_chg = nday_pct_chg[:(0 - n_steps)]
nday_pct_chg = nday_pct_chg.values
y = nday_pct_chg.reshape((len(nday_pct_chg), 1))
y = to_categorical(y)
#print(y)
dataset = series.values
X = split_sequences(dataset, n_steps)
#print(X.shape, y.shape)
train_len = int(0.88 * len(X))
train_X = X[:train_len, :]
test_X = X[train_len:, :]
train_y = y[:train_len, :]
test_y = y[train_len:, :]
#print(train_X.shape, train_y.shape, test_X.shape, test_y.shape)
return train_X, train_y, test_X, test_y
# fit and evaluate a model
def evaluate_model(trainX, trainy, testX, testy):
verbose, epochs, batch_size = 0, 15, 32
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[2], trainy.shape[1]
model = Sequential()
#model.add(LSTM(10, input_shape=(n_timesteps, n_features), return_sequences=True))
#model.add(LSTM(10, input_shape=(n_timesteps, n_features), return_sequences=True))
model.add(LSTM(10, input_shape=(n_timesteps, n_features)))
model.add(Dense(10, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
# fit network
model.fit(trainX, trainy, epochs=epochs, batch_size=batch_size, verbose=verbose)
# evaluate model
_, accuracy = model.evaluate(testX, testy, batch_size=batch_size, verbose=0)
return accuracy
# summarize scores
def summarize_results(scores):
print(scores)
m, s = mean(scores), std(scores)
print('Accuracy: %.3f%% (+/-%.3f)' % (m, s))
# run an experiment
def run_experiment(repeats=1):
for ndays in range(1, 2):
for nsteps in range(10, 20):
# load data
trainX, trainy, testX, testy = load_dataset(ndays, nsteps)
# repeat experiment
scores = list()
for r in range(repeats):
score = evaluate_model(trainX, trainy, testX, testy)
score = score * 100.0
#print('>#%d: %.3f' % (r + 1, score))
scores.append(score)
print('ndays: %d, nsteps: %d, score: %.3f' % (ndays, nsteps, mean(scores)))
if repeats > 1:
# summarize results
summarize_results(scores)
# run the experiment
run_experiment()
Raw
logicbelief-master-puwork_ml_daily.py
import tushare as ts
import math
import warnings
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
# Obtain a database connection to the MySQL instance
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
data = pro.stock_basic(exchange_id='', list_status='L', fields='ts_code')
ts_code = data['ts_code']
api = ts.pro_api()
for stock in ts_code:
df = ts.pro_bar(pro_api = api, ts_code = stock, adj = 'qfq', start_date = '20130101', end_date = '20191231', ma=[5, 20, 50, 200])
if df is None:
continue
#print(df)
cur = con.cursor()
column_str = """date, code, open, close, high, low, volume, pct_chg, ma5, ma20, ma50, ma200, v_ma5, v_ma20, v_ma50, v_ma200, day_price_chg, week_price_chg, month_price_chg"""
for i in range(0, len(df)):
if i < 20:
continue
row = df.iloc[i]
if math.isnan(row['ma50']) or math.isnan(row['ma200']):
continue
day_price_chg = (df.iloc[i-1]['close']/row['close'] - 1) * 100
week_price_chg = (df.iloc[i-5]['close']/row['close'] - 1) * 100
month_price_chg = (df.iloc[i-20]['close']/row['close'] - 1) * 100
final_str = "REPLACE INTO daily (%s) VALUES ('%s', '%s', '%s', '%s','%s', '%s', '%s', '%s','%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', '%s', '%s', '%s')" % \
(column_str,
row['trade_date'],
row['ts_code'],
row['open'],
row['close'],
row['high'],
row['low'],
row['vol'],
row['pct_chg'],
row['ma5'],
row['ma20'],
row['ma50'],
row['ma200'],
row['ma_v_5'],
row['ma_v_20'],
row['ma_v_50'],
row['ma_v_200'],
day_price_chg,
week_price_chg,
month_price_chg)
cur.execute(final_str)
con.commit()
print('%s finish' %(stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_daily_basic.py
from hurst import *
from common import *
import math
import warnings
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
with open("whole") as f:
content = f.readlines()
whole = [x.strip() for x in content]
for stock in whole:
df = None
while True:
try:
df = pro.daily_basic(ts_code=stock,
fields='trade_date, ts_code, turnover_rate_f,volume_ratio,pe,pb,ps,circ_mv, total_mv',
start_date='20130801', end_date='20191231')
break
except:
print('%s error' % stock)
time.sleep(5)
if df is None:
continue
cur = con.cursor()
for i in range(0, len(df) - 1):
row = df.iloc[i]
try:
if math.isnan(row['turnover_rate_f']) or \
row['volume_ratio'] is None or \
math.isnan(row['volume_ratio']) or \
math.isnan(row['pe']) or \
math.isnan(row['pb']) or \
math.isnan(row['circ_mv']) or \
math.isnan(row['total_mv']) or \
math.isnan(row['ps']):
continue
except:
continue
final_str = "UPDATE daily SET turnover_rate_f={}, volume_ratio={}, pe={}, pb={}, ps={}, circ_mv={}, total_mv={} WHERE date = '{}' AND code = '{}'".format(
row['turnover_rate_f'],
row['volume_ratio'],
row['pe'],
row['pb'],
row['ps'],
row['circ_mv'],
row['total_mv'],
row['trade_date'],
row['ts_code'])
# print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_dnn.py
import numpy as np
from sklearn.preprocessing import MinMaxScaler # library required for scaling the data
import pandas as pd
import warnings
warnings.filterwarnings("ignore")
data = pd.read_csv('nn/Recurrent Neural Networks/SBIN.csv',index_col = 0,parse_dates = True, na_values = 0).dropna()
max_=data[['Open','High','Low','Close']].max().max()
min_=data[['Open','High','Low','Close']].min().min()
X1=(data[['Open','High','Low','Close']]-min_)/(max_-min_)
X1=np.array(X1)
data[['Volume']] = data[['Volume']].astype(float)
scl=MinMaxScaler()
X2=scl.fit_transform(data[['Volume']].values.reshape(-1,1))
data=data.assign(Open=X1[:,0])
data=data.assign(High=X1[:,1])
data=data.assign(Low=X1[:,2])
data=data.assign(Close=X1[:,3])
data=data.assign(Volume=X2[:,0])
X=data[['Open','High','Low','Close','Volume']]
y=np.ones(len(data))
y=np.where(data.Close.shift(-5)>data.Close,1,0)
test_size=100
X_train=np.array(X.iloc[:-test_size])
y_train=np.array(y[:-test_size])
X_test=np.array(X.iloc[-test_size:])
y_test=np.array(y[-test_size:])
class_1=y_train.sum()
class_0=len(y_train)-class_1
class_0_percentage=class_0/len(y_train)
class_1_percentage=class_1/len(y_train)
class_0_percentage=class_0/len(y_train)
class_1_percentage=class_1/len(y_train)
class_0_weight=class_1_percentage
class_1_weight=class_0_percentage
class_weight={0:class_0_weight,1:class_1_weight}
# Libraries
from keras.models import Sequential
from keras.layers import Dropout
from keras.layers import Dense, Activation
from keras.callbacks import ModelCheckpoint
from keras.layers import BatchNormalization
# Parameters
dropout_ratio=0.20
momentum=0.99
np.random.seed(42)
neurons=175
act_1='tanh'
act_2='softmax'
# The DNN Model
print(X_train.shape[1:])
model=Sequential() #We want the layers to work sequentially, one after other.
# Please Type Your Code Below
model.add(Dense(neurons, use_bias=True, kernel_initializer='he_normal', bias_initializer='zeros',input_shape=X_train.shape[1:]))
Raw
logicbelief-master-puwork_ml_predict_short.py
from datetime import datetime
from hurst import *
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import requests
import MySQLdb as mdb
url = 'http://pai-eas-vpc.cn-shanghai.aliyuncs.com/api/predict/short'
headers = {"Authorization": 'MDU4NDgwOWNjNjY5ODY1NmFiMmE4MzRmN2EzZjYzNTVhMGVhMDgzMw=='}
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock, list_date, industry, market, is_hs, name from stock_basic"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], datetime.combine(d[1], datetime.min.time()), d[2], d[3], d[4], d[5]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
predict_results = []
date = 0
start_date = '20190901'
end_date = '20201231'
for (stock, list_date, industry, market, is_hs, name) in obtain_stock_basic():
if 'ST' in name or name.startswith('688'):
continue
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date=start_date, end_date=end_date, ma=[5, 10, 20, 60])
if df is None or len(df) < 60:
continue
if df.head(60)['amount'].mean() < 30000:
continue
# print(df)
for i in range(0, 1):
row = df.iloc[i]
if row['trade_date'] is None:
continue
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
if row.pct_chg > 9.9:
continue
# print(row['trade_date'])
last_60_days = df[i: i + 60][::-1]
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_v_ratio_v2(last_10_days)
udvr20 = up_down_v_ratio_v2(last_20_days)
udvr30 = up_down_v_ratio_v2(last_30_days)
udvr60 = up_down_v_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
cur = con.cursor()
data = '[{"md10":%s, "md20":%d, "md30":%d, "md60":%s, "mp10":%s, "mp20":%s, \
"mp30":%s, "mp60":%s, "udvr10":%s, "udvr20":%s, "udvr30":%s, "udvr60":%s, "udpr10":%s, "udpr20":%s, \
"udpr30":%s, "udpr60":%s, "av1":%s, "av3":%s, "av5":%s, "av7":%s, "av10":%s, "av20":%s, "av60":%s, \
"pc1":%s, "pc3":%s, "pc5":%s, "pc7":%s, "pc10":%s, "pc20":%s, "pc60":%s,\
"pr_5c":%s, "pr_510":%s, "pr_1020":%s, "pr_2060":%s,\
"vr_5c":%s, "vr_510":%s, "vr_1020":%s, "vr_2060":%s }]' % \
(md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060)
# print(data)
resp = requests.post(url, data=data, headers=headers)
if resp.status_code != 200:
print(resp.content)
else:
#print(resp.content)
value = float(str(resp.content).split(":")[1].split("}")[0])
if value > 5:
query = "'%s', '%s', '%s'" % (row['trade_date'], stock, value)
predict_results.append(query)
print(query)
print('%s finish' % (stock))
print(predict_results)
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
# Obtain a database connection to the MySQL instance
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
cur = con.cursor()
for q in predict_results:
query = "INSERT IGNORE INTO predict_short (date, code, value) VALUES (%s)" % (q)
print(query)
cur.execute(query)
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_predict_short_hk.py
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import requests
import MySQLdb as mdb
url = 'http://pai-eas-vpc.cn-shanghai.aliyuncs.com/api/predict/short_hk'
headers = {"Authorization": 'Y2E3YWEzYzM3ZThjZGFlYTdlZDc2NmQwNjQwMTBmZjg0NTkyZjk5MQ=='}
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_hk_basic where amount > 300000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
predict_results = []
date = 0
start_date = '20190901'
end_date = '20201231'
for stock in obtain_stock_basic():
try:
df = pro.hk_daily(ts_code=stock,start_date=start_date, end_date=end_date, ma=[5, 10, 20, 60])
except:
print("%s error" % stock)
if df is None or len(df) < 60:
continue
#print(df)
rdf = df[::-1]
rdf = sma(rdf, "ma5", n=5)
rdf = sma(rdf, "ma10", n=10)
rdf = sma(rdf, "ma20", n=20)
rdf = sma(rdf, "ma60", n=60)
rdf = sma(rdf, "ma_v_5", n=5, val_name='vol')
rdf = sma(rdf, "ma_v_10", n=10, val_name='vol')
rdf = sma(rdf, "ma_v_20", n=20, val_name='vol')
rdf = sma(rdf, "ma_v_60", n=60, val_name='vol')
df = rdf[::-1]
for i in range(0, 1):
row = df.iloc[i]
#print(row)
if row['trade_date'] is None:
continue
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
last_60_days = df[i: i + 60][::-1]
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_v_ratio_v2(last_10_days)
udvr20 = up_down_v_ratio_v2(last_20_days)
udvr30 = up_down_v_ratio_v2(last_30_days)
udvr60 = up_down_v_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
cur = con.cursor()
data = '[{"md10":%s, "md20":%d, "md30":%d, "md60":%s, "mp10":%s, "mp20":%s, \
"mp30":%s, "mp60":%s, "udvr10":%s, "udvr20":%s, "udvr30":%s, "udvr60":%s, "udpr10":%s, "udpr20":%s, \
"udpr30":%s, "udpr60":%s, "av1":%s, "av3":%s, "av5":%s, "av7":%s, "av10":%s, "av20":%s, "av60":%s, \
"pc1":%s, "pc3":%s, "pc5":%s, "pc7":%s, "pc10":%s, "pc20":%s, "pc60":%s,\
"pr_5c":%s, "pr_510":%s, "pr_1020":%s, "pr_2060":%s,\
"vr_5c":%s, "vr_510":%s, "vr_1020":%s, "vr_2060":%s }]' % \
(md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060)
#print(data)
resp = requests.post(url, data=data, headers=headers)
if resp.status_code != 200:
print(resp.content)
else:
#print(resp.content)
value = float(str(resp.content).split(":")[1].split("}")[0])
if value > 7 or value < -5:
query = "'%s', '%s', '%s'" % (row['trade_date'], stock, value)
predict_results.append(query)
print(query)
print('%s finish' % (stock))
print(predict_results)
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
# Obtain a database connection to the MySQL instance
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
cur = con.cursor()
for q in predict_results:
query = "INSERT IGNORE INTO predict_short_hk (date, code, value) VALUES (%s)" % (q)
print(query)
cur.execute(query)
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_predict_short_us.py
import warnings
from factor import *
import pandas_datareader.data as web
from datetime import datetime
warnings.simplefilter(action='ignore')
import requests
import MySQLdb as mdb
url = 'http://pai-eas-vpc.cn-shanghai.aliyuncs.com/api/predict/short_us'
headers = {"Authorization": 'OGNjNjQyNjlmNjFlMzEzMTIwNTc5NjY0MWE3MGU3Y2QxZmNiMzExZA=='}
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_us limit 1000"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
predict_results = []
date = 0
start = datetime(2019, 10, 1)
end = datetime(2020, 12, 31)
for stock in obtain_stock_basic():
try:
df = web.DataReader(stock, 'iex', start, end, access_key='pk_88d98ab1d0344ceb8184b898313a18cc')
if df is None or len(df) < 60:
continue
df['pct_chg'] = df['close'].pct_change()
df = sma(df, "ma5", n=5)
df = sma(df, "ma10", n=10)
df = sma(df, "ma20", n=20)
df = sma(df, "ma60", n=60)
df = sma(df, "ma_v_5", n=5, val_name='volume')
df = sma(df, "ma_v_10", n=10, val_name='volume')
df = sma(df, "ma_v_20", n=20, val_name='volume')
df = sma(df, "ma_v_60", n=60, val_name='volume')
for i in range(len(df) - 1, len(df)):
row = df.iloc[i]
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
last_60_days = df.tail(60)
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_volume_ratio_v2(last_10_days)
udvr20 = up_down_volume_ratio_v2(last_20_days)
udvr30 = up_down_volume_ratio_v2(last_30_days)
udvr60 = up_down_volume_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['volume']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
cur = con.cursor()
data = '[{"md10":%s, "md20":%d, "md30":%d, "md60":%s, "mp10":%s, "mp20":%s, \
"mp30":%s, "mp60":%s, "udvr10":%s, "udvr20":%s, "udvr30":%s, "udvr60":%s, "udpr10":%s, "udpr20":%s, \
"udpr30":%s, "udpr60":%s, "av1":%s, "av3":%s, "av5":%s, "av7":%s, "av10":%s, "av20":%s, "av60":%s, \
"pc1":%s, "pc3":%s, "pc5":%s, "pc7":%s, "pc10":%s, "pc20":%s, "pc60":%s,\
"pr_5c":%s, "pr_510":%s, "pr_1020":%s, "pr_2060":%s,\
"vr_5c":%s, "vr_510":%s, "vr_1020":%s, "vr_2060":%s }]' % \
(md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060)
print(data)
resp = requests.post(url, data=data, headers=headers)
if resp.status_code != 200:
print(resp.content)
else:
# print(resp.content)
value = float(str(resp.content).split(":")[1].split("}")[0])
if value > 2 or value < -1:
query = "'%s', '%s', '%s'" % (row.name, stock, value)
predict_results.append(query)
print(query)
print('%s finish' % (stock))
except Exception as e:
print(e)
print('%s error' % (stock))
print(predict_results)
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
# Obtain a database connection to the MySQL instance
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
cur = con.cursor()
for q in predict_results:
query = "INSERT IGNORE INTO predict_short_us (date, code, value) VALUES (%s)" % (q)
print(query)
cur.execute(query)
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_predict_test.py
import requests
url = 'http://pai-eas-vpc.cn-shanghai.aliyuncs.com/api/predict/daily_lb'
headers = {"Authorization": 'MGQ1MDgxNTExYjllZTg2MDk3YmM4MWQwMzhkNTdjZDZlZWRmNTMwOA=='}
data = '[{"open":7.63, "close":7.43, "high":7.79, "low":7.43, "volume":122927, "pct_chg":-4.12903225806, "ma5":7.92, "ma20":8.84, "ma50":7.04,"ma200":5.79, "v_ma5":152690.88, "v_ma20":357717.28, "v_ma50":249808.42, "v_ma200":78657.53, "turnover_rate_f":6.4754, "circ_mv":305770.8073, "total_mv":306844.27, "volume_ratio":0.72, "pe":81.3303, "pb":3.1623, "ps":3}]'
resp = requests.post(url, data=data, headers=headers)
if resp.status_code != 200:
print(resp.content)
else:
print(resp.content)
print(str(resp.content).split(":")[1].split("}")[0])
print(str(resp.content).split(":}"))
Raw
logicbelief-master-puwork_ml_predict_test_m.py
import requests
url = 'http://pai-eas-vpc.cn-shanghai.aliyuncs.com/api/predict/daily_lb'
headers = {"Authorization": 'MGQ1MDgxNTExYjllZTg2MDk3YmM4MWQwMzhkNTdjZDZlZWRmNTMwOA=='}
url_month = 'http://pai-eas-vpc.cn-shanghai.aliyuncs.com/api/predict/month'
headers_month = {"Authorization": 'MzM2MGRiMzQ5ZmM0OWUzZTZhMzYzMjE4ZWM4ZGUwN2MxMDA5MjQ0YQ=='}
data = '[{"open":7.63, "close":7.43, "high":7.79, "low":7.43, "volume":122927, "pct_chg":-4.12903225806, "ma5":7.92, "ma20":8.84, "ma50":7.04,"ma200":5.79, "v_ma5":152690.88, "v_ma20":357717.28, "v_ma50":249808.42, "v_ma200":78657.53, "turnover_rate_f":6.4754, "circ_mv":305770.8073, "total_mv":306844.27, "volume_ratio":0.72, "pe":81.3303, "pb":3.1623, "ps":3}]'
#resp = requests.post(url, data=data, headers=headers)
resp = requests.post(url_month, data=data, headers=headers_month)
if resp.status_code != 200:
print(resp.content)
else:
print(resp.content)
print(str(resp.content).split(":")[1].split("}")[0])
print(str(resp.content).split(":}"))
Raw
logicbelief-master-puwork_ml_sepa_daily.py
from datetime import datetime
from hurst import *
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock, list_date, industry, market, is_hs from stock_basic"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], datetime.combine(d[1], datetime.min.time()), d[2], d[3], d[4]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
start_date = '20130101'
end_date = '20191231'
for (stock, list_date, industry, market, is_hs) in obtain_stock_basic():
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date = start_date, end_date = end_date, ma=[50, 150, 200])
if df is None or len(df) < 260:
continue
# print(df)
baisc_df = None
while True:
try:
basic_df = pro.daily_basic(ts_code=stock,
fields='trade_date, turnover_rate_f,volume_ratio,pe,pe_ttm, pb,ps,circ_mv, total_mv',
start_date=start_date, end_date=end_date)
break
except:
print('%s error' % stock)
time.sleep(5)
if basic_df is None or len(basic_df) == 0:
continue
basic_df = basic_df.set_index('trade_date')
for i in range(15, len(df) - 260):
row = df.iloc[i]
if row['trade_date'] is None:
continue
# print(row)
if math.isnan(row['ma200']) or math.isnan(row['ma50']) or math.isnan(row['ma150']):
continue
basic_row = basic_df.loc[row['trade_date']]
#print(basic_row)
if basic_row['pe'] is None or \
basic_row['pb'] is None or \
basic_row['ps'] is None or \
math.isnan(basic_row['pe']) or \
math.isnan(basic_row['pe_ttm']) or \
math.isnan(basic_row['pb']) or math.isnan(basic_row['ps']):
continue
trade_date = datetime.strptime(row['trade_date'], "%Y%m%d")
list_months = int((trade_date - list_date).days / 30)
ma50 = row['ma50']
ma150 = row['ma150']
ma200 = row['ma200']
v_ma50 = row['ma_v_50']
v_ma150 = row['ma_v_150']
v_ma200 = row['ma_v_200']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
price_50_150_ratio = ma50 / ma150
price_150_close_ratio = ma150 / close
price_200_close_ratio = ma200 / close
v_50_volume_ratio = v_ma50 / volume
v_200_volume_ratio = v_ma200 / volume
v_50_150_ratio = v_ma50 / v_ma150
volatility = (row['high'] / row['low'] - 1) * 100
df_30 = df[i: i + 30]
up_days_200 = days_above(df_30, 'ma200')
up_days_50 = days_above(df_30, 'ma50')
df_260 = df[i: i + 260]
pct_below_highest = 1 - close / df_260['close'].max()
pct_up_lowest = close / df_260['close'].min() - 1
df_last_week = df[i: i + 5]
up_days_last_week = up_days(df_last_week)
close_up_days_last_week = close_up_days(df_last_week)
df_last_two_week = df[i: i + 10]
up_days_last_two_week = up_days(df_last_two_week)
close_up_days_last_two_week = close_up_days(df_last_two_week)
up_down_v_ratio_last_two_week = up_down_v_ratio(df_last_two_week)
df_last_15_days = df[i: i + 15]
up_days_last_15_days = up_days(df_last_15_days)
pct_chg_last_15_days = df_last_15_days['pct_chg'].mean()
up_down_v_ratio_last_15_days = up_down_v_ratio(df_last_15_days)
days_15_avg_volatility = avg_volatility(df_last_15_days)
price_chg_15_days = ((df.iloc[i-15]['close'] / close) - 1) * 100
cur = con.cursor()
column_str = """date, code, list_month, market, is_hs, 50_150_ratio, 150_close_ratio, 200_close_ratio, 50_volume_ratio, pct_chg, volatility, 200_up_days, 50_up_days, pct_below_highest, pct_up_lowest, industry, \
200_volume_ratio, 50_150_v_ratio, turnover_rate_f, circ_mv, total_mv, volume_ratio, pe, pe_ttm, pb, ps, \
up_days_last_week, up_days_last_two_week, close_up_days_last_week, close_up_days_last_two_week,\
up_down_v_ratio_last_two_week, up_down_v_ratio_last_15_days, pct_chg_last_15_days, up_days_last_15_days, 15days_avg_volatility, 15days_price_chg"""
final_str = "REPLACE INTO sepa (%s) VALUES ('%s', '%s', '%s', '%s', '%s', '%s','%s', '%s', '%s', '%s','%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s')" % \
(column_str,
row['trade_date'],
stock,
list_months,
market,
is_hs,
price_50_150_ratio,
price_150_close_ratio,
price_200_close_ratio,
v_50_volume_ratio,
pct_chg,
volatility,
up_days_200,
up_days_50,
pct_below_highest,
pct_up_lowest,
industry,
v_200_volume_ratio,
v_50_150_ratio,
basic_row['turnover_rate_f'],
basic_row['circ_mv'],
basic_row['total_mv'],
basic_row['volume_ratio'],
basic_row['pe'],
basic_row['pe_ttm'],
basic_row['pb'],
basic_row['ps'],
up_days_last_week,
up_days_last_two_week,
close_up_days_last_week,
close_up_days_last_two_week,
up_down_v_ratio_last_two_week,
up_down_v_ratio_last_15_days,
pct_chg_last_15_days,
up_days_last_15_days,
days_15_avg_volatility,
price_chg_15_days)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_sepa_update.py
from datetime import datetime
from hurst import *
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock, list_date, industry, market, is_hs from stock_basic WHERE stock LIKE '%SZ'"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], datetime.combine(d[1], datetime.min.time()), d[2], d[3], d[4]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
start_date = '20130101'
end_date = '20191231'
for (stock, list_date, industry, market, is_hs) in obtain_stock_basic():
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date = start_date, end_date = end_date, ma=[50, 150, 200])
if df is None or len(df) < 260:
continue
#print(df)
for i in range(15, len(df) - 260):
row = df.iloc[i]
if row['trade_date'] is None:
continue
if math.isnan(row['ma200']) or math.isnan(row['ma50']) or math.isnan(row['ma150']):
continue
#print(row['trade_date'])
trade_date = datetime.strptime(row['trade_date'], "%Y%m%d")
new_year_date = trade_date.replace(month=1, day=1)
days = (trade_date - new_year_date).days
#print("%s %s %d %s" %(trade_date, new_year_date, days, row['trade_date']))
list_months = int((trade_date - list_date).days / 30)
df_last_5_days = df[i: i + 5]
df_last_10_days = df[i: i + 10]
df_last_15_days = df[i: i + 15]
df_last_30_days = df[i: i + 30]
df_last_60_days = df[i: i + 60]
up_down_v_ratio_last_two_week = up_down_v_ratio_v2(df_last_10_days)
up_down_v_ratio_last_15_days = up_down_v_ratio_v2(df_last_15_days)
up_down_v_ratio_last_30_days = up_down_v_ratio_v2(df_last_30_days)
days_5_avg_volatility = avg_volatility(df_last_5_days)
days_15_avg_volatility = avg_volatility(df_last_15_days)
days_30_avg_volatility = avg_volatility(df_last_30_days)
days_60_avg_volatility = avg_volatility(df_last_60_days)
up_down_ratio_last_10_days = up_down_pct_chg_ratio(df_last_10_days)
up_down_ratio_last_15_days = up_down_pct_chg_ratio(df_last_15_days)
up_down_ratio_last_30_days = up_down_pct_chg_ratio(df_last_30_days)
cur = con.cursor()
query = " UPDATE sepa SET days = {}, up_down_v_ratio_last_two_week = {}, up_down_v_ratio_last_15_days = {}, up_down_v_ratio_last_30_days = {}, " \
"5days_avg_volatility = {}, 15days_avg_volatility = {}, 30days_avg_volatility = {}, 60days_avg_volatility = {}, " \
"up_down_ratio_last_10_days = {}, up_down_ratio_last_15_days = {}, up_down_ratio_last_30_days = {}, list_month= {} WHERE date = '{}' AND code = '{}'"\
.format(days, up_down_v_ratio_last_two_week, up_down_v_ratio_last_15_days, up_down_v_ratio_last_30_days,
days_5_avg_volatility, days_15_avg_volatility, days_30_avg_volatility, days_60_avg_volatility,
up_down_ratio_last_10_days, up_down_ratio_last_15_days, up_down_ratio_last_30_days, list_months, row.trade_date, stock)
#print(query)
cur.execute(query)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_short_hk.py
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_hk_basic where amount > 30000000 limit 160, 500"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
start_date = '20170101'
end_date = '20191231'
for stock in obtain_stock_basic():
try:
df = pro.hk_daily(ts_code=stock,start_date=start_date, end_date=end_date, ma=[5, 10, 20, 60])
except:
print("%s error" % stock)
if df is None or len(df) < 60:
continue
df['ma5'] = ma(df, n=5)
df['ma10'] = ma(df, n=10)
df['ma20'] = ma(df, n=20)
df['ma60'] = ma(df, n=60)
df['ma_v_5'] = ma(df, n=5, val_name='vol')
df['ma_v_10'] = ma(df, n=10, val_name='vol')
df['ma_v_20'] = ma(df, n=20, val_name='vol')
df['ma_v_60'] = ma(df, n=60, val_name='vol')
#print(df.head(10))
for i in range(6, len(df) - 60):
row = df.iloc[i]
if row['trade_date'] is None:
continue
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
# print(row['trade_date'])
last_60_days = df[i: i + 60][::-1]
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_v_ratio_v2(last_10_days)
udvr20 = up_down_v_ratio_v2(last_20_days)
udvr30 = up_down_v_ratio_v2(last_30_days)
udvr60 = up_down_v_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
target = round((df.iloc[i - 6]['open'] / df.iloc[i - 1]['open'] - 1) * 100, 2)
cur = con.cursor()
column_str = """date, code, md10,md20,md30,md60,mp10,mp20,mp30,mp60,\
udvr10,udvr20,udvr30,udvr60,udpr10,udpr20,udpr30,udpr60,\
av1,av3,av5,av7,av10,av20,av60,pc1,pc3,pc5,pc7,pc10,pc20,pc60,\
pr_5c,pr_510,pr_1020,pr_2060,vr_5c,vr_510,vr_1020,vr_2060,target"""
final_str = "REPLACE INTO short_hk (%s) VALUES ('%s', '%s', '%s', '%s', '%s', '%s','%s', '%s', '%s', '%s',\
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s', '%s', '%s', '%s')" % \
(column_str,
row['trade_date'],
stock,
md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060,
target)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_short_model.py
from datetime import datetime
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock, list_date, industry, market, is_hs from stock_basic"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0], datetime.combine(d[1], datetime.min.time()), d[2], d[3], d[4]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
start_date = '20170101'
end_date = '20191231'
for (stock, list_date, industry, market, is_hs) in obtain_stock_basic():
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date=start_date, end_date=end_date, ma=[5, 10, 20, 60])
if df is None or len(df) < 60:
continue
# print(df)
for i in range(470, len(df) - 60):
row = df.iloc[i]
if row['trade_date'] is None:
continue
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
# print(row['trade_date'])
last_60_days = df[i: i + 60][::-1]
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_v_ratio_v2(last_10_days)
udvr20 = up_down_v_ratio_v2(last_20_days)
udvr30 = up_down_v_ratio_v2(last_30_days)
udvr60 = up_down_v_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
target = round((df.iloc[i - 4]['open'] / df.iloc[i - 1]['open'] - 1) * 100, 2)
cur = con.cursor()
column_str = """date, code, md10,md20,md30,md60,mp10,mp20,mp30,mp60,\
udvr10,udvr20,udvr30,udvr60,udpr10,udpr20,udpr30,udpr60,\
av1,av3,av5,av7,av10,av20,av60,pc1,pc3,pc5,pc7,pc10,pc20,pc60,\
pr_5c,pr_510,pr_1020,pr_2060,vr_5c,vr_510,vr_1020,vr_2060,target"""
final_str = "REPLACE INTO short (%s) VALUES ('%s', '%s', '%s', '%s', '%s', '%s','%s', '%s', '%s', '%s',\
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s', '%s', '%s', '%s')" % \
(column_str,
row['trade_date'],
stock,
md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060,
target)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_short_model_hk.py
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_hk_basic where amount > 30000000 limit 0, 160"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
start_date = '20170101'
end_date = '20191231'
for stock in obtain_stock_basic():
try:
df = pro.hk_daily(ts_code=stock,start_date=start_date, end_date=end_date, ma=[5, 10, 20, 60])
except:
print("%s error" % stock)
if df is None or len(df) < 60:
continue
rdf = df[::-1]
rdf = sma(rdf, "ma5", n=5)
rdf = sma(rdf, "ma10", n=10)
rdf = sma(rdf, "ma20", n=20)
rdf = sma(rdf, "ma60", n=60)
rdf = sma(rdf, "ma_v_5", n=5, val_name='vol')
rdf = sma(rdf, "ma_v_10", n=10, val_name='vol')
rdf = sma(rdf, "ma_v_20", n=20, val_name='vol')
rdf = sma(rdf, "ma_v_60", n=60, val_name='vol')
df = rdf[::-1]
#print(df.head(10))
for i in range(6, len(df) - 60):
row = df.iloc[i]
if row['trade_date'] is None:
continue
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
# print(row['trade_date'])
last_60_days = df[i: i + 60][::-1]
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_v_ratio_v2(last_10_days)
udvr20 = up_down_v_ratio_v2(last_20_days)
udvr30 = up_down_v_ratio_v2(last_30_days)
udvr60 = up_down_v_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
target = round((df.iloc[i - 6]['open'] / df.iloc[i - 1]['open'] - 1) * 100, 2)
cur = con.cursor()
column_str = """date, code, md10,md20,md30,md60,mp10,mp20,mp30,mp60,\
udvr10,udvr20,udvr30,udvr60,udpr10,udpr20,udpr30,udpr60,\
av1,av3,av5,av7,av10,av20,av60,pc1,pc3,pc5,pc7,pc10,pc20,pc60,\
pr_5c,pr_510,pr_1020,pr_2060,vr_5c,vr_510,vr_1020,vr_2060,target"""
final_str = "REPLACE INTO short_hk (%s) VALUES ('%s', '%s', '%s', '%s', '%s', '%s','%s', '%s', '%s', '%s',\
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s', '%s', '%s', '%s')" % \
(column_str,
row['trade_date'],
stock,
md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060,
target)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_short_model_us.py
import warnings
from factor import *
import pandas_datareader.data as web
from datetime import datetime
import math
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_us limit 0, 261"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
start = datetime(2017, 1, 1)
end = datetime(2020, 1, 21)
for stock in obtain_stock_basic():
try:
df = web.DataReader(stock, 'iex', start, end, access_key='pk_88d98ab1d0344ceb8184b898313a18cc')
if df is None or len(df) < 60:
continue
df['pct_chg'] = df['close'].pct_change()
df = sma(df, "ma5", n=5)
df = sma(df, "ma10", n=10)
df = sma(df, "ma20", n=20)
df = sma(df, "ma60", n=60)
df = sma(df, "ma_v_5", n=5, val_name='volume')
df = sma(df, "ma_v_10", n=10, val_name='volume')
df = sma(df, "ma_v_20", n=20, val_name='volume')
df = sma(df, "ma_v_60", n=60, val_name='volume')
for i in range(60, len(df) - 6):
row = df.iloc[i]
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
last_60_days = df[i-59: i+1]
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_volume_ratio_v2(last_10_days)
udvr20 = up_down_volume_ratio_v2(last_20_days)
udvr30 = up_down_volume_ratio_v2(last_30_days)
udvr60 = up_down_volume_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['volume']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
target = round((df.iloc[i+6]['open'] / df.iloc[i+1]['open'] - 1) * 100, 2)
cur = con.cursor()
column_str = """date, code, md10,md20,md30,md60,mp10,mp20,mp30,mp60,\
udvr10,udvr20,udvr30,udvr60,udpr10,udpr20,udpr30,udpr60,\
av1,av3,av5,av7,av10,av20,av60,pc1,pc3,pc5,pc7,pc10,pc20,pc60,\
pr_5c,pr_510,pr_1020,pr_2060,vr_5c,vr_510,vr_1020,vr_2060,target"""
final_str = "REPLACE INTO short_us (%s) VALUES ('%s', '%s', '%s', '%s', '%s', '%s','%s', '%s', '%s', '%s',\
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s', '%s', '%s', '%s')" % \
(column_str,
row.name,
stock,
md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060,
target)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
except Exception as e:
print(e)
print('%s error' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_short_us.py
import warnings
from factor import *
import pandas_datareader.data as web
from datetime import datetime
import math
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_us limit 261, 500"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
start = datetime(2017, 1, 1)
end = datetime(2020, 1, 21)
for stock in obtain_stock_basic():
try:
df = web.DataReader(stock, 'iex', start, end, access_key='pk_88d98ab1d0344ceb8184b898313a18cc')
if df is None or len(df) < 60:
continue
df['pct_chg'] = df['close'].pct_change()
df = sma(df, "ma5", n=5)
df = sma(df, "ma10", n=10)
df = sma(df, "ma20", n=20)
df = sma(df, "ma60", n=60)
df = sma(df, "ma_v_5", n=5, val_name='volume')
df = sma(df, "ma_v_10", n=10, val_name='volume')
df = sma(df, "ma_v_20", n=20, val_name='volume')
df = sma(df, "ma_v_60", n=60, val_name='volume')
for i in range(60, len(df) - 6):
row = df.iloc[i]
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
last_60_days = df[i-59: i+1]
last_30_days = last_60_days.tail(30)
last_20_days = last_30_days.tail(20)
last_15_days = last_20_days.tail(15)
last_10_days = last_15_days.tail(10)
last_7_days = last_10_days.tail(7)
last_5_days = last_10_days.tail(5)
last_3_days = last_10_days.tail(3)
last_day = last_3_days.tail(1)
md10 = max_drawback(last_10_days)
md20 = max_drawback(last_20_days)
md30 = max_drawback(last_30_days)
md60 = max_drawback(last_60_days)
mp10 = max_profit(last_10_days)
mp20 = max_profit(last_20_days)
mp30 = max_profit(last_30_days)
mp60 = max_profit(last_60_days)
udvr10 = up_down_volume_ratio_v2(last_10_days)
udvr20 = up_down_volume_ratio_v2(last_20_days)
udvr30 = up_down_volume_ratio_v2(last_30_days)
udvr60 = up_down_volume_ratio_v2(last_60_days)
udpr10 = up_down_pct_chg_ratio(last_10_days)
udpr20 = up_down_pct_chg_ratio(last_20_days)
udpr30 = up_down_pct_chg_ratio(last_30_days)
udpr60 = up_down_pct_chg_ratio(last_60_days)
av1 = avg_volatility(last_day)
av3 = avg_volatility(last_3_days)
av5 = avg_volatility(last_5_days)
av7 = avg_volatility(last_7_days)
av10 = avg_volatility(last_10_days)
av20 = avg_volatility(last_20_days)
av60 = avg_volatility(last_60_days)
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['volume']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
pc1 = avg_pct_chg(last_day)
pc3 = avg_pct_chg(last_3_days)
pc5 = avg_pct_chg(last_5_days)
pc7 = avg_pct_chg(last_7_days)
pc10 = avg_pct_chg(last_10_days)
pc20 = avg_pct_chg(last_20_days)
pc60 = avg_pct_chg(last_60_days)
target = round((df.iloc[i+6]['open'] / df.iloc[i+1]['open'] - 1) * 100, 2)
cur = con.cursor()
column_str = """date, code, md10,md20,md30,md60,mp10,mp20,mp30,mp60,\
udvr10,udvr20,udvr30,udvr60,udpr10,udpr20,udpr30,udpr60,\
av1,av3,av5,av7,av10,av20,av60,pc1,pc3,pc5,pc7,pc10,pc20,pc60,\
pr_5c,pr_510,pr_1020,pr_2060,vr_5c,vr_510,vr_1020,vr_2060,target"""
final_str = "REPLACE INTO short_us (%s) VALUES ('%s', '%s', '%s', '%s', '%s', '%s','%s', '%s', '%s', '%s',\
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s','%s', '%s', '%s', '%s', \
'%s', '%s', '%s', '%s', '%s', '%s', '%s')" % \
(column_str,
row.name,
stock,
md10,
md20,
md30,
md60,
mp10,
mp20,
mp30,
mp60,
udvr10,
udvr20,
udvr30,
udvr60,
udpr10,
udpr20,
udpr30,
udpr60,
av1,
av3,
av5,
av7,
av10,
av20,
av60,
pc1,
pc3,
pc5,
pc7,
pc10,
pc20,
pc60,
pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060,
target)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
except Exception as e:
print(e)
print('%s error' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_update_hk.py
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_hk_basic where amount > 30000000 limit 160, 200"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
start_date = '20170101'
end_date = '20191231'
for stock in obtain_stock_basic():
try:
df = pro.hk_daily(ts_code=stock,start_date=start_date, end_date=end_date, ma=[5, 10, 20, 60])
except:
print("%s error" % stock)
if df is None or len(df) < 60:
continue
rdf = df[::-1]
rdf = sma(rdf, "ma5", n=5)
rdf = sma(rdf, "ma10", n=10)
rdf = sma(rdf, "ma20", n=20)
rdf = sma(rdf, "ma60", n=60)
rdf = sma(rdf, "ma_v_5", n=5, val_name='vol')
rdf = sma(rdf, "ma_v_10", n=10, val_name='vol')
rdf = sma(rdf, "ma_v_20", n=20, val_name='vol')
rdf = sma(rdf, "ma_v_60", n=60, val_name='vol')
df = rdf[::-1]
#print(df.head(10))
for i in range(6, len(df) - 60):
row = df.iloc[i]
if row['trade_date'] is None:
continue
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
cur = con.cursor()
final_str = "UPDATE short_hk set pr_5c='%s', pr_510='%s', pr_1020='%s', pr_2060='%s', \
vr_5c='%s', vr_510='%s', vr_1020='%s', vr_2060='%s' where date='%s' and code='%s'" % \
(pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060,
row['trade_date'],
stock)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_ml_update_model_hk.py
import tushare as ts
import warnings
from factor import *
warnings.simplefilter(action='ignore')
import MySQLdb as mdb
with open("db") as f:
content = f.readlines()
content = [x.strip() for x in content]
db_host = content[0]
db_user = content[1]
db_pass = content[2]
db_name = content[3]
con = mdb.connect(db_host, db_user, db_pass, db_name, port=3306, charset="utf8", use_unicode=True)
def obtain_stock_basic():
cur = con.cursor()
query_str = "select stock from stock_hk_basic where amount > 30000000 limit 0, 160"
cur.execute(query_str)
data = cur.fetchall()
cur.close()
return [(d[0]) for d in data]
ts.set_token('d014b0f46a93f4d9472162c809649eb34b0e9770b8a6d0de3155df04')
pro = ts.pro_api()
start_date = '20170101'
end_date = '20191231'
for stock in obtain_stock_basic():
try:
df = pro.hk_daily(ts_code=stock,start_date=start_date, end_date=end_date, ma=[5, 10, 20, 60])
except:
print("%s error" % stock)
if df is None or len(df) < 60:
continue
rdf = df[::-1]
rdf = sma(rdf, "ma5", n=5)
rdf = sma(rdf, "ma10", n=10)
rdf = sma(rdf, "ma20", n=20)
rdf = sma(rdf, "ma60", n=60)
rdf = sma(rdf, "ma_v_5", n=5, val_name='vol')
rdf = sma(rdf, "ma_v_10", n=10, val_name='vol')
rdf = sma(rdf, "ma_v_20", n=20, val_name='vol')
rdf = sma(rdf, "ma_v_60", n=60, val_name='vol')
df = rdf[::-1]
#print(df.head(10))
for i in range(6, len(df) - 60):
row = df.iloc[i]
if row['trade_date'] is None:
continue
if math.isnan(row['ma60']) or math.isnan(row['ma20']) or math.isnan(row['ma10']):
continue
ma5 = row['ma5']
ma10 = row['ma10']
ma20 = row['ma20']
ma60 = row['ma60']
v_ma5 = row['ma_v_5']
v_ma10 = row['ma_v_10']
v_ma20 = row['ma_v_20']
v_ma60 = row['ma_v_60']
close = row['close']
volume = row['vol']
pct_chg = row['pct_chg']
pr_5c = ma5 / close
pr_510 = ma5 / ma10
pr_1020 = ma10 / ma20
pr_2060 = ma20 / ma60
vr_5c = v_ma5 / volume
vr_510 = v_ma5 / v_ma10
vr_1020 = v_ma10 / v_ma20
vr_2060 = v_ma20 / v_ma60
cur = con.cursor()
final_str = "UPDATE short_hk set pr_5c='%s', pr_510='%s', pr_1020='%s', pr_2060='%s', \
vr_5c='%s', vr_510='%s', vr_1020='%s', vr_2060='%s' where date='%s' and code='%s'" % \
(pr_5c,
pr_510,
pr_1020,
pr_2060,
vr_5c,
vr_510,
vr_1020,
vr_2060,
row['trade_date'],
stock)
#print(final_str)
cur.execute(final_str)
con.commit()
print('%s finish' % (stock))
con.commit()
con.close()
print("Done")
Raw
logicbelief-master-puwork_new.py
from common import *
import warnings
import math
warnings.simplefilter(action='ignore')
def is_cross(df):
for index, row in df.iterrows():
if math.fabs(row['open'] / row['close'] - 1) < 0.005 and \
row['vol'] / row['ma_v_5'] < 0.9 and \
row['close'] / row['ma5'] < 1.014:
print(row['trade_date'])
return True
return False
stocks = pro.stock_basic(exchange_id='', list_status='L', fields='ts_code, name, list_date')
start_date = (datetime.today() - timedelta(days=120)).strftime("%Y%m%d")
end_date = (datetime.today() - timedelta(days=30)).strftime("%Y%m%d")
column_str = """date, code, ocr, v5r, c5r"""
queries = []
for row in stocks.itertuples():
if 'ST' in row[2]:
continue
# if not('688' in row[1]):
# continue
if row[3] < start_date or row[3] > end_date:
continue
df = ts.pro_bar(api=pro, ts_code=row[1], adj='qfq', start_date=start_date, end_date='20220423', ma=[5, 10])
if is_cross(df.head(1)):
print("%s" %(row[1]))
last = df.head(1)
ocr = math.fabs(last['open'].max() / last['close'].max() - 1) * 100
v5r = last['vol'].max() / last['ma_v_5'].max()
c5r = last['close'] / last['ma5'].max()
query = "INSERT IGNORE INTO new (%s) VALUES (%s, '%s', %.2f, %.2f, %.2f)" % \
(column_str, df.head(1)['trade_date'].max(), row[1], ocr, v5r, c5r)
queries.append(query)
print('%s finish' % (row[1]))
con = get_conn("db8")
execute_queries(con, queries)
print("Done")
Raw
logicbelief-master-puwork_new_high_low.py
from common import *
import warnings
import datetime
warnings.simplefilter(action='ignore', category=FutureWarning)
data = pro.stock_basic(exchange_id='', list_status='L', fields='ts_code,name,list_date')
with open("1800") as f:
content = f.readlines()
content = [x.strip() for x in content]
stocks_300 = set(content)
stock_len = len(stocks_300)
count = 0
new_high = 0
new_low = 0
date = 0
for row in data.itertuples():
listdate = int(row[3])
if listdate > 20190101:
continue
stock = row[1]
name = row[2]
code = stock.split('.', 1)
if not (code[0] in stocks_300):
continue
count += 1
if count % 50 == 0:
print('finish %d/%d' % (count, stock_len))
try:
last_year = datetime.date.today() - datetime.timedelta(days=365)
start_date = last_year.strftime("%Y%m%d")
df = ts.pro_bar(api=pro, ts_code=stock, adj='qfq', start_date=start_date, end_date='20201230')
except Exception as e:
continue
d = int(df.head(1)['trade_date'].max())
if d > date:
date = d
last_close = df.head(1)['close'].max()
max_close = df['close'].max()
min_close = df['close'].min()
if last_close >= max_close:
new_high += 1
elif last_close <= min_close:
new_low += 1
print('high/low %d/%d' % (new_high, new_low))
con = get_conn()
query = format("INSERT IGNORE INTO high_vs_low (date, high, low) VALUES (%d,%d,%d)" % (date, new_high, new_low))
execute_query(con, query)
print("Done")
Raw
logicbelief-master-puwork_nn_Challenges in Live Trading_BAC_Daily.csv
date open high low close volume
2011-01-03 00:00:00+00:00 14.19 14.23 13.8 13.85 354322300
2011-01-04 00:00:00+00:00 14.24 14.25 14.02 14.23 218978200
2011-01-05 00:00:00+00:00 14.5 14.6 14.15 14.19 246151200
2011-01-06 00:00:00+00:00 14.44 14.69 14.34 14.54 241658500
2011-01-07 00:00:00+00:00 14.25 14.68 13.98 14.54 392328700
2011-01-10 00:00:00+00:00 14.4 14.43 14.09 14.17 185382600
2011-01-11 00:00:00+00:00 14.69 14.73 14.53 14.61 212239400
2011-01-12 00:00:00+00:00 14.99 14.99 14.85 14.89 204060800
2011-01-13 00:00:00+00:00 14.77 15.02 14.72 15.01 158944400
2011-01-14 00:00:00+00:00 15.25 15.31 14.68 14.735 282493800
2011-01-18 00:00:00+00:00 15 15.16 14.85 15.08 198400500
2011-01-19 00:00:00+00:00 14.37 14.95 14.35 14.85 247013200
2011-01-20 00:00:00+00:00 14.54 14.605 13.94 14.27 245219600
2011-01-21 00:00:00+00:00 14.25 14.71 14.22 14.41 291812600
2011-01-24 00:00:00+00:00 13.92 14.26 13.88 14.25 225424700
2011-01-25 00:00:00+00:00 13.63 13.84 13.4 13.78 303642100
2011-01-26 00:00:00+00:00 13.55 13.77 13.55 13.71 145810300
2011-01-27 00:00:00+00:00 13.67 13.67 13.48 13.58 153086100
2011-01-28 00:00:00+00:00 13.6 14.06 13.58 13.83 226452500
2011-01-31 00:00:00+00:00 13.73 13.79 13.64 13.71 118000800
2011-02-01 00:00:00+00:00 14.31 14.37 13.87 13.895 211978400
2011-02-02 00:00:00+00:00 14.24 14.35 14.13 14.33 140312700
2011-02-03 00:00:00+00:00 14.43 14.47 14.15 14.16 145885300
2011-02-04 00:00:00+00:00 14.29 14.47 14.11 14.43 141015200
2011-02-07 00:00:00+00:00 14.67 14.77 14.43 14.51 149423500
2011-02-08 00:00:00+00:00 14.61 14.76 14.5 14.64 158426200
2011-02-09 00:00:00+00:00 14.64 14.69 14.41 14.46 150179100
2011-02-10 00:00:00+00:00 14.49 14.64 14.47 14.505 132240800
2011-02-11 00:00:00+00:00 14.77 14.87 14.35 14.37 156195600
2011-02-14 00:00:00+00:00 14.89 14.95 14.71 14.775 112571100
2011-02-15 00:00:00+00:00 14.77 14.88 14.69 14.8 109532300
2011-02-16 00:00:00+00:00 14.84 14.88 14.7 14.81 132821900
2011-02-17 00:00:00+00:00 14.81 14.91 14.73 14.75 103495400
2011-02-18 00:00:00+00:00 14.75 14.89 14.67 14.84 98369000
2011-02-22 00:00:00+00:00 14.18 14.52 14.09 14.38 187522800
2011-02-23 00:00:00+00:00 14.17 14.44 13.92 14.17 196355700
2011-02-24 00:00:00+00:00 13.97 14.16 13.79 14.11 201697200
2011-02-25 00:00:00+00:00 14.2 14.32 14.12 14.16 126869100
2011-02-28 00:00:00+00:00 14.29 14.48 14.16 14.27 137039100
2011-03-01 00:00:00+00:00 13.93 14.35 13.91 14.315 161263900
2011-03-02 00:00:00+00:00 13.83 14.07 13.81 13.92 115609400
2011-03-03 00:00:00+00:00 14.27 14.29 14.05 14.05 139513500
2011-03-04 00:00:00+00:00 14.12 14.31 13.98 14.3 146184700
2011-03-07 00:00:00+00:00 14.03 14.27 13.92 14.18 139010300
2011-03-08 00:00:00+00:00 14.69 14.7 14.2 14.27 250033200
2011-03-09 00:00:00+00:00 14.59 14.69 14.48 14.66 148290400
2011-03-10 00:00:00+00:00 14.26 14.46 14.26 14.42 155462700
2011-03-11 00:00:00+00:00 14.38 14.43 14.1 14.11 111555000
2011-03-14 00:00:00+00:00 14.23 14.35 14.07 14.26 112139700
2011-03-15 00:00:00+00:00 13.96 14.06 13.71 13.77 167837200
2011-03-16 00:00:00+00:00 13.69 14.1 13.66 14.01 178630900
2011-03-17 00:00:00+00:00 13.98 14.04 13.75 13.89 131745300
2011-03-18 00:00:00+00:00 14.04 14.29 13.98 14.2 199209100
2011-03-21 00:00:00+00:00 14.05 14.22 13.9 14.2 114391500
2011-03-22 00:00:00+00:00 13.88 14.05 13.88 14.04 86286500
2011-03-23 00:00:00+00:00 13.65 13.74 13.37 13.72 230626600
2011-03-24 00:00:00+00:00 13.48 13.59 13.32 13.56 170761100
2011-03-25 00:00:00+00:00 13.34 13.53 13.32 13.49 115322600
2011-03-28 00:00:00+00:00 13.37 13.56 13.37 13.42 78640400
2011-03-29 00:00:00+00:00 13.35 13.41 13.16 13.41 117737500
2011-03-30 00:00:00+00:00 13.45 13.56 13.27 13.4 121082200
2011-03-31 00:00:00+00:00 13.33 13.39 13.29 13.35 86180900
2011-04-01 00:00:00+00:00 13.37 13.61 13.35 13.45 95017100
2011-04-04 00:00:00+00:00 13.44 13.59 13.4 13.4 71113900
2011-04-05 00:00:00+00:00 13.47 13.5 13.37 13.43 64968800
2011-04-06 00:00:00+00:00 13.72 13.78 13.53 13.6 136151600
2011-04-07 00:00:00+00:00 13.61 13.88 13.54 13.79 119206000
2011-04-08 00:00:00+00:00 13.48 13.72 13.45 13.63 87592400
2011-04-11 00:00:00+00:00 13.49 13.59 13.43 13.5 63348300
2011-04-12 00:00:00+00:00 13.47 13.58 13.31 13.4 100372600
2011-04-13 00:00:00+00:00 13.27 13.64 13.21 13.61 124630800
2011-04-14 00:00:00+00:00 13.13 13.26 13.07 13.16 115624900
2011-04-15 00:00:00+00:00 12.82 13.33 12.82 13.22 266336200
2011-04-18 00:00:00+00:00 12.42 12.6 12.24 12.59 262012500
2011-04-19 00:00:00+00:00 12.34 12.53 12.15 12.53 182612900
2011-04-20 00:00:00+00:00 12.27 12.48 12.21 12.43 179989100
2011-04-21 00:00:00+00:00 12.31 12.4 12.24 12.33 100820900
2011-04-25 00:00:00+00:00 12.44 12.6 12.32 12.325 112414300
2011-04-26 00:00:00+00:00 12.23 12.53 12.23 12.49 146800000
2011-04-27 00:00:00+00:00 12.33 12.4 12.23 12.24 118705100
2011-04-28 00:00:00+00:00 12.42 12.45 12.25 12.31 92414200
2011-04-29 00:00:00+00:00 12.28 12.42 12.28 12.39 87169300
2011-05-02 00:00:00+00:00 12.34 12.47 12.33 12.36 89940100
2011-05-03 00:00:00+00:00 12.6 12.71 12.33 12.35 145530300
2011-05-04 00:00:00+00:00 12.49 12.68 12.42 12.64 99234300
2011-05-05 00:00:00+00:00 12.3 12.47 12.2 12.41 130921200
2011-05-06 00:00:00+00:00 12.31 12.45 12.28 12.41 111324100
2011-05-09 00:00:00+00:00 12.18 12.315 12.11 12.28 110668200
2011-05-10 00:00:00+00:00 12.28 12.37 12.18 12.21 91577700
2011-05-11 00:00:00+00:00 12.25 12.43 12.23 12.275 150517800
2011-05-12 00:00:00+00:00 12.2 12.27 12.09 12.2 111941900
2011-05-13 00:00:00+00:00 11.93 12.21 11.91 12.195 159056700
2011-05-16 00:00:00+00:00 11.86 12.11 11.82 11.89 123646400
2011-05-17 00:00:00+00:00 11.9 11.94 11.8 11.81 145394700
2011-05-18 00:00:00+00:00 11.79 11.9 11.73 11.9 152963400
2011-05-19 00:00:00+00:00 11.69 11.89 11.68 11.88 121082700
2011-05-20 00:00:00+00:00 11.58 11.78 11.53 11.64 115821500
2011-05-23 00:00:00+00:00 11.42 11.52 11.38 11.47 122134100
2011-05-24 00:00:00+00:00 11.46 11.51 11.29 11.45 129513400
2011-05-25 00:00:00+00:00 11.38 11.58 11.35 11.35 103583400
2011-05-26 00:00:00+00:00 11.46 11.56 11.38 11.39 124072200
2011-05-27 00:00:00+00:00 11.69 11.78 11.54 11.57 107370400
2011-05-31 00:00:00+00:00 11.75 11.92 11.62 11.87 106228200
2011-06-01 00:00:00+00:00 11.24 11.74 11.23 11.71 176331800
2011-06-02 00:00:00+00:00 11.29 11.42 11.08 11.29 198074000
2011-06-03 00:00:00+00:00 11.28 11.45 11.16 11.2 115786700
2011-06-06 00:00:00+00:00 10.83 11.2 10.75 11.18 213289400
2011-06-07 00:00:00+00:00 10.65 11.05 10.6 10.96 160224800
2011-06-08 00:00:00+00:00 10.54 10.79 10.495 10.59 159384200
2011-06-09 00:00:00+00:00 10.65 10.75 10.5 10.58 129533300
2011-06-10 00:00:00+00:00 10.8 11.03 10.41 10.62 210796300
2011-06-13 00:00:00+00:00 10.97 11.02 10.64 10.89 186095900
2011-06-14 00:00:00+00:00 10.8 11.12 10.75 11.12 173051000
2011-06-15 00:00:00+00:00 10.5 10.67 10.41 10.6 213408200
2011-06-16 00:00:00+00:00 10.6 10.69 10.4 10.45 200127400
2011-06-17 00:00:00+00:00 10.68 10.77 10.58 10.75 116367000
2011-06-20 00:00:00+00:00 10.6 10.66 10.53 10.59 88402500
2011-06-21 00:00:00+00:00 10.83 10.87 10.53 10.68 109392600
2011-06-22 00:00:00+00:00 10.79 10.94 10.78 10.79 102448600
2011-06-23 00:00:00+00:00 10.71 10.77 10.56 10.65 180604400
2011-06-24 00:00:00+00:00 10.52 10.77 10.485 10.73 122210200
2011-06-27 00:00:00+00:00 10.85 10.95 10.52 10.52 202814500
2011-06-28 00:00:00+00:00 10.82 10.92 10.77 10.91 113027000
2011-06-29 00:00:00+00:00 11.14 11.25 11 11.17 301628700
2011-06-30 00:00:00+00:00 10.96 11.18 10.84 11.17 255993900
2011-07-01 00:00:00+00:00 11.09 11.14 10.92 10.98 137487700
2011-07-05 00:00:00+00:00 11 11.07 10.91 11.06 110261100
2011-07-06 00:00:00+00:00 10.74 10.85 10.66 10.83 145455700
2011-07-07 00:00:00+00:00 10.92 10.96 10.85 10.9 113940400
2011-07-08 00:00:00+00:00 10.7 10.8 10.61 10.75 129514100
2011-07-11 00:00:00+00:00 10.35 10.55 10.3 10.53 145323900
2011-07-12 00:00:00+00:00 10.21 10.4 10.2 10.27 145156100
2011-07-13 00:00:00+00:00 10.2 10.36 10.15 10.27 164765500
2011-07-14 00:00:00+00:00 10.07 10.33 10.06 10.3 148536100
2011-07-15 00:00:00+00:00 10 10.18 9.88 10.12 183416000
2011-07-18 00:00:00+00:00 9.72 9.93 9.53 9.88 226934200
2011-07-19 00:00:00+00:00 9.57 9.85 9.4 9.75 322430300
2011-07-20 00:00:00+00:00 9.85 10 9.63 9.66 249783800
2011-07-21 00:00:00+00:00 10.23 10.28 10 10.03 238464700
2011-07-22 00:00:00+00:00 10.13 10.27 10.05 10.26 138441500
2011-07-25 00:00:00+00:00 10.01 10.07 9.88 9.97 130582700
2011-07-26 00:00:00+00:00 10 10.09 9.91 9.97 121168700
2011-07-27 00:00:00+00:00 9.68 9.94 9.68 9.92 151491700
2011-07-28 00:00:00+00:00 9.79 9.85 9.69 9.72 113623300
2011-07-29 00:00:00+00:00 9.71 9.95 9.6 9.64 187926700
2011-08-01 00:00:00+00:00 9.81 10.05 9.7 10.04 159616700
2011-08-02 00:00:00+00:00 9.49 9.86 9.47 9.76 173972600
2011-08-03 00:00:00+00:00 9.54 9.59 9.32 9.49 175691200
2011-08-04 00:00:00+00:00 8.83 9.46 8.77 9.37 305309300
2011-08-05 00:00:00+00:00 8.17 9.05 8.029 8.97 546781600
2011-08-08 00:00:00+00:00 6.51 7.7 6.31 7.4 682052900
2011-08-09 00:00:00+00:00 7.6 7.66 6.68 7.02 496094500
2011-08-10 00:00:00+00:00 6.77 7.59 6.74 7.53 493986500
2011-08-11 00:00:00+00:00 7.25 7.42 6.96 7.1 344079000
2011-08-12 00:00:00+00:00 7.19 7.71 7.13 7.49 298997800
2011-08-15 00:00:00+00:00 7.76 7.84 7.35 7.46 268370400
2011-08-16 00:00:00+00:00 7.4 7.66 7.34 7.61 275997300
2011-08-17 00:00:00+00:00 7.46 7.59 7.4 7.46 159346200
2011-08-18 00:00:00+00:00 7.01 7.09 6.78 7.05 335205500
2011-08-19 00:00:00+00:00 6.97 7.12 6.75 6.8 289217900
2011-08-22 00:00:00+00:00 6.42 7.22 6.42 7.2 398777800
2011-08-23 00:00:00+00:00 6.3 6.44 6.01 6.37 563788800
2011-08-24 00:00:00+00:00 6.99 7.05 6.32 6.36 602753000
2011-08-25 00:00:00+00:00 7.65 8.8 7.38 8.29 859643400
2011-08-26 00:00:00+00:00 7.76 7.98 7.45 7.62 424194600
2011-08-29 00:00:00+00:00 8.39 8.41 7.96 8.1 363420500
2011-08-30 00:00:00+00:00 8.119 8.39 8.05 8.29 297220400
2011-08-31 00:00:00+00:00 8.17 8.29 8.109 8.28 282263500
2011-09-01 00:00:00+00:00 7.91 8.18 7.91 8.18 243549600
2011-09-02 00:00:00+00:00 7.25 7.45 7.17 7.31 305988600
2011-09-06 00:00:00+00:00 6.99 7.07 6.8 6.91 283101800
2011-09-07 00:00:00+00:00 7.48 7.52 7.1 7.23 257251800
2011-09-08 00:00:00+00:00 7.2 7.44 7.2 7.38 204593600
2011-09-09 00:00:00+00:00 6.98 7.21 6.96 7.11 255828800
2011-09-12 00:00:00+00:00 7.05 7.1 6.81 7 286856000
2011-09-13 00:00:00+00:00 7 7.18 6.97 7.09 220657500
2011-09-14 00:00:00+00:00 7.05 7.13 6.92 7.09 212940200
2011-09-15 00:00:00+00:00 7.33 7.34 7.13 7.18 230299700
2011-09-16 00:00:00+00:00 7.23 7.39 7.08 7.38 240523700
2011-09-19 00:00:00+00:00 6.99 7.07 6.935 7.06 199254700
2011-09-20 00:00:00+00:00 6.9 7.06 6.9 7.04 142828700
2011-09-21 00:00:00+00:00 6.38 6.97 6.36 6.95 381953600
2011-09-22 00:00:00+00:00 6.06 6.28 6 6.24 384622500
2011-09-23 00:00:00+00:00 6.31 6.39 6.07 6.09 503374000
2011-09-26 00:00:00+00:00 6.6 6.6 6.31 6.48 228406400
2011-09-27 00:00:00+00:00 6.48 6.85 6.41 6.84 231598900
2011-09-28 00:00:00+00:00 6.16 6.53 6.16 6.51 222449300
2011-09-29 00:00:00+00:00 6.35 6.45 6.15 6.4 224013100
2011-09-30 00:00:00+00:00 6.12 6.315 6.11 6.18 175213100
2011-10-03 00:00:00+00:00 5.53 6.18 5.52 6.08 369335100
2011-10-04 00:00:00+00:00 5.76 5.76 5.13 5.48 448300000
2011-10-05 00:00:00+00:00 5.77 5.83 5.51 5.71 291785700
2011-10-06 00:00:00+00:00 6.28 6.31 5.65 5.77 336087400
2011-10-07 00:00:00+00:00 5.9 6.33 5.88 6.32 285598300
2011-10-10 00:00:00+00:00 6.28 6.29 6.12 6.14 225326600
2011-10-11 00:00:00+00:00 6.37 6.47 6.13 6.18 204431100
2011-10-12 00:00:00+00:00 6.58 6.74 6.34 6.51 284617700
2011-10-13 00:00:00+00:00 6.22 6.44 6.17 6.44 230115400
2011-10-14 00:00:00+00:00 6.19 6.36 6.12 6.31 203564500
2011-10-17 00:00:00+00:00 6.03 6.28 6.03 6.18 175512200
2011-10-18 00:00:00+00:00 6.64 6.79 6.16 6.27 496883800
2011-10-19 00:00:00+00:00 6.4 6.86 6.37 6.68 318554300
2011-10-20 00:00:00+00:00 6.47 6.48 6.18 6.43 254971700
2011-10-21 00:00:00+00:00 6.46 6.65 6.38 6.59 252586400
2011-10-24 00:00:00+00:00 6.72 6.74 6.49 6.59 217032400
2011-10-25 00:00:00+00:00 6.46 6.67 6.46 6.65 200837500
2011-10-26 00:00:00+00:00 6.59 6.66 6.44 6.58 208197400
2011-10-27 00:00:00+00:00 7.05 7.23 6.9 7.11 408959900
2011-10-28 00:00:00+00:00 7.35 7.43 7.05 7.08 273167300
2011-10-31 00:00:00+00:00 6.83 7.16 6.82 7.09 252026900
2011-11-01 00:00:00+00:00 6.4 6.68 6.32 6.38 371823200
2011-11-02 00:00:00+00:00 6.72 6.74 6.57 6.72 216830700
2011-11-03 00:00:00+00:00 6.91 6.98 6.58 6.91 244340400
2011-11-04 00:00:00+00:00 6.49 6.75 6.45 6.73 266446300
2011-11-07 00:00:00+00:00 6.45 6.59 6.298 6.44 192433500
2011-11-08 00:00:00+00:00 6.53 6.59 6.4 6.54 219000600
2011-11-09 00:00:00+00:00 6.16 6.36 6.15 6.31 265539000
2011-11-10 00:00:00+00:00 6.03 6.33 6.01 6.32 324988000
2011-11-11 00:00:00+00:00 6.21 6.29 6.12 6.12 210025900
2011-11-14 00:00:00+00:00 6.05 6.19 6.02 6.16 225481700
2011-11-15 00:00:00+00:00 6.13 6.24 6 6.01 268228500
2011-11-16 00:00:00+00:00 5.9 6.09 5.9 6.09 292502400
2011-11-17 00:00:00+00:00 5.8 6.03 5.79 5.98 285881900
2011-11-18 00:00:00+00:00 5.78 5.89 5.75 5.86 227038600
2011-11-21 00:00:00+00:00 5.49 5.68 5.48 5.66 269658300
2011-11-22 00:00:00+00:00 5.37 5.585 5.32 5.52 270959000
2011-11-23 00:00:00+00:00 5.14 5.31 5.13 5.3 264978500
2011-11-25 00:00:00+00:00 5.17 5.33 5.12 5.16 138425100
2011-11-28 00:00:00+00:00 5.25 5.53 5.14 5.5 345537900
2011-11-29 00:00:00+00:00 5.08 5.28 5.03 5.19 333749700
2011-11-30 00:00:00+00:00 5.44 5.44 5.18 5.4 436268200
2011-12-01 00:00:00+00:00 5.53 5.63 5.29 5.37 315495500
2011-12-02 00:00:00+00:00 5.64 5.88 5.61 5.67 283132900
2011-12-05 00:00:00+00:00 5.79 5.95 5.73 5.86 293977700
2011-12-06 00:00:00+00:00 5.78 5.92 5.7 5.74 254787200
2011-12-07 00:00:00+00:00 5.89 5.92 5.7 5.73 244700100
2011-12-08 00:00:00+00:00 5.59 5.88 5.53 5.84 286031100
2011-12-09 00:00:00+00:00 5.72 5.85 5.63 5.67 293654100
2011-12-12 00:00:00+00:00 5.45 5.6 5.4 5.59 191824000
2011-12-13 00:00:00+00:00 5.32 5.6 5.22 5.56 227194200
2011-12-14 00:00:00+00:00 5.23 5.37 5.2 5.24 226723100
2011-12-15 00:00:00+00:00 5.26 5.37 5.213 5.33 196092100
2011-12-16 00:00:00+00:00 5.2 5.42 5.16 5.31 227111300
2011-12-19 00:00:00+00:00 4.985 5.22 4.92 5.2 344317700
2011-12-20 00:00:00+00:00 5.17 5.22 5.04 5.11 239246100
2011-12-21 00:00:00+00:00 5.23 5.25 5.1 5.18 203711800
2011-12-22 00:00:00+00:00 5.47 5.51 5.21 5.26 303085900
2011-12-23 00:00:00+00:00 5.6 5.63 5.47 5.54 190712000
2011-12-27 00:00:00+00:00 5.48 5.58 5.46 5.55 158575400
2011-12-28 00:00:00+00:00 5.285 5.46 5.27 5.45 146104500
2011-12-29 00:00:00+00:00 5.46 5.48 5.28 5.29 168098500
2011-12-30 00:00:00+00:00 5.56 5.58 5.37 5.39 176441000
2012-01-03 00:00:00+00:00 5.8 5.89 5.74 5.75 246293200
2012-01-04 00:00:00+00:00 5.81 5.88 5.62 5.71 243711200
2012-01-05 00:00:00+00:00 6.31 6.35 5.71 5.75 547780000
2012-01-06 00:00:00+00:00 6.18 6.3 6.06 6.21 299630600
2012-01-09 00:00:00+00:00 6.27 6.37 6.19 6.26 240614400
2012-01-10 00:00:00+00:00 6.63 6.66 6.44 6.44 354292100
2012-01-11 00:00:00+00:00 6.87 6.9 6.52 6.6 354039600
2012-01-12 00:00:00+00:00 6.79 7.02 6.66 6.99 362099600
2012-01-13 00:00:00+00:00 6.61 6.69 6.41 6.49 337048400
2012-01-17 00:00:00+00:00 6.48 6.82 6.44 6.63 294196500
2012-01-18 00:00:00+00:00 6.8 6.8 6.46 6.5 302497700
2012-01-19 00:00:00+00:00 6.96 7.29 6.745 7.21 491008100
2012-01-20 00:00:00+00:00 7.07 7.08 6.83 6.95 236630700
2012-01-23 00:00:00+00:00 7.25 7.37 7.11 7.13 339942200
2012-01-24 00:00:00+00:00 7.29 7.35 7.06 7.11 228810300
2012-01-25 00:00:00+00:00 7.35 7.37 7.15 7.2 249340100
2012-01-26 00:00:00+00:00 7.3 7.5 7.23 7.45 265938800
2012-01-27 00:00:00+00:00 7.29 7.35 7.2 7.21 230914000
2012-01-30 00:00:00+00:00 7.07 7.15 7.02 7.13 231240100
2012-01-31 00:00:00+00:00 7.13 7.19 7.05 7.17 212736300
2012-02-01 00:00:00+00:00 7.36 7.44 7.21 7.25 318811000
2012-02-02 00:00:00+00:00 7.45 7.49 7.33 7.43 232425400
2012-02-03 00:00:00+00:00 7.84 7.89 7.63 7.66 364195700
2012-02-06 00:00:00+00:00 7.97 7.97 7.77 7.785 236316000
2012-02-07 00:00:00+00:00 7.85 7.99 7.8 7.95 253694200
2012-02-08 00:00:00+00:00 8.13 8.15 7.93 7.96 436439700
2012-02-09 00:00:00+00:00 8.18 8.35 8.1 8.31 480206700
2012-02-10 00:00:00+00:00 8.07 8.119 7.98 8.05 254420800
2012-02-13 00:00:00+00:00 8.25 8.31 8.21 8.27 308490200
2012-02-14 00:00:00+00:00 7.98 8.17 7.95 8.15 384682800
2012-02-15 00:00:00+00:00 7.78 8.119 7.77 8.01 372981400
2012-02-16 00:00:00+00:00 8.09 8.13 7.66 7.71 337131800
2012-02-17 00:00:00+00:00 8.02 8.08 7.92 8.039 355445600
2012-02-21 00:00:00+00:00 8.109 8.2 8 8.025 333407700
2012-02-22 00:00:00+00:00 7.95 8.13 7.95 8.06 228867800
2012-02-23 00:00:00+00:00 8.02 8.05 7.9 7.97 144231800
2012-02-24 00:00:00+00:00 7.88 8.09 7.88 8.05 163128200
2012-02-27 00:00:00+00:00 8.039 8.05 7.67 7.79 298930800
2012-02-28 00:00:00+00:00 8.119 8.15 8 8.05 193443200
2012-02-29 00:00:00+00:00 7.97 8.24 7.97 8.16 266720500
2012-03-01 00:00:00+00:00 8.119 8.17 8.05 8.09 197850100
2012-03-02 00:00:00+00:00 8.13 8.21 8.095 8.109 143887500
2012-03-05 00:00:00+00:00 7.97 8.13 7.95 8.09 196328300
2012-03-06 00:00:00+00:00 7.71 7.79 7.66 7.78 262871200
2012-03-07 00:00:00+00:00 8.02 8.02 7.77 7.81 328331900
2012-03-08 00:00:00+00:00 8.06 8.119 8 8.07 160282000
2012-03-09 00:00:00+00:00 8.05 8.189 8.01 8.115 195281400
2012-03-12 00:00:00+00:00 7.99 8.06 7.91 8.01 165298900
2012-03-13 00:00:00+00:00 8.49 8.5 8.05 8.07 385984100
2012-03-14 00:00:00+00:00 8.84 8.9 8.56 8.66 488077000
2012-03-15 00:00:00+00:00 9.24 9.25 8.85 8.98 489593000
2012-03-16 00:00:00+00:00 9.8 9.8 9.33 9.41 582294100
2012-03-19 00:00:00+00:00 9.53 10.1 9.51 9.78 669479900
2012-03-20 00:00:00+00:00 9.81 9.97 9.59 9.63 451213900
2012-03-21 00:00:00+00:00 9.82 10.03 9.738 9.96 326573800
2012-03-22 00:00:00+00:00 9.6 9.77 9.52 9.66 264520500
2012-03-23 00:00:00+00:00 9.85 9.85 9.4 9.47 283110900
2012-03-26 00:00:00+00:00 9.93 10.04 9.85 10.01 241605700
2012-03-27 00:00:00+00:00 9.6 9.9 9.58 9.865 250069600
2012-03-28 00:00:00+00:00 9.75 9.78 9.58 9.62 234021500
2012-03-29 00:00:00+00:00 9.53 9.68 9.43 9.63 253625800
2012-03-30 00:00:00+00:00 9.57 9.64 9.35 9.61 250234500
2012-04-02 00:00:00+00:00 9.68 9.78 9.41 9.54 179229000
2012-04-03 00:00:00+00:00 9.49 9.67 9.41 9.67 190453400
2012-04-04 00:00:00+00:00 9.2 9.36 9.15 9.35 228560600
2012-04-05 00:00:00+00:00 9.23 9.4 9.11 9.14 180868100
2012-04-09 00:00:00+00:00 8.93 9.04 8.83 9.04 211427600
2012-04-10 00:00:00+00:00 8.54 9.09 8.5 8.97 377220000
2012-04-11 00:00:00+00:00 8.86 8.91 8.72 8.78 252563700
2012-04-12 00:00:00+00:00 9.17 9.18 8.91 8.93 223979900
2012-04-13 00:00:00+00:00 8.68 9.08 8.68 9.08 282354600
2012-04-16 00:00:00+00:00 8.79 8.93 8.62 8.87 217284500
2012-04-17 00:00:00+00:00 8.92 9 8.87 8.965 193424400
2012-04-18 00:00:00+00:00 8.92 9 8.84 8.88 160064100
2012-04-19 00:00:00+00:00 8.77 9.17 8.67 9.16 348804500
2012-04-20 00:00:00+00:00 8.36 8.78 8.33 8.78 277965300
2012-04-23 00:00:00+00:00 8.18 8.3 7.95 8.02 256073900
2012-04-24 00:00:00+00:00 8.21 8.27 8.1 8.25 192443800
2012-04-25 00:00:00+00:00 8.26 8.35 8.17 8.3 164902400
2012-04-26 00:00:00+00:00 8.27 8.4 8.17 8.189 131586300
2012-04-27 00:00:00+00:00 8.25 8.34 8.189 8.34 120948200
2012-04-30 00:00:00+00:00 8.109 8.24 8.039 8.22 137422300
2012-05-01 00:00:00+00:00 8.31 8.4 8.08 8.109 177953400
2012-05-02 00:00:00+00:00 8.16 8.23 8.119 8.21 136208900
2012-05-03 00:00:00+00:00 8 8.18 7.91 8.175 204584500
2012-05-04 00:00:00+00:00 7.74 7.92 7.7 7.9 194857200
2012-05-07 00:00:00+00:00 7.96 8 7.66 7.71 167662600
2012-05-08 00:00:00+00:00 7.79 7.9 7.69 7.87 180627900
2012-05-09 00:00:00+00:00 7.73 7.82 7.65 7.67 174742300
2012-05-10 00:00:00+00:00 7.7 7.94 7.67 7.91 156787300
2012-05-11 00:00:00+00:00 7.55 7.77 7.41 7.48 245533600
2012-05-14 00:00:00+00:00 7.35 7.52 7.35 7.37 170716800
2012-05-15 00:00:00+00:00 7.3 7.55 7.25 7.37 201254900
2012-05-16 00:00:00+00:00 7.11 7.47 7.08 7.4 216226100
2012-05-17 00:00:00+00:00 6.98 7.25 6.93 7.1 241468300
2012-05-18 00:00:00+00:00 7.02 7.03 6.89 7.01 223783500
2012-05-21 00:00:00+00:00 6.83 7.11 6.72 7.03 229980800
2012-05-22 00:00:00+00:00 6.98 7.15 6.86 6.92 221551500
2012-05-23 00:00:00+00:00 7.17 7.17 6.85 6.9 194441500
2012-05-24 00:00:00+00:00 7.14 7.33 7.01 7.26 206041500
2012-05-25 00:00:00+00:00 7.15 7.25 7.07 7.07 108661200
2012-05-29 00:00:00+00:00 7.44 7.45 7.22 7.28 159872200
2012-05-30 00:00:00+00:00 7.2 7.41 7.1 7.33 206801500
2012-05-31 00:00:00+00:00 7.35 7.37 7.04 7.21 203267300
2012-06-01 00:00:00+00:00 7.02 7.19 6.94 7.12 240859000
2012-06-04 00:00:00+00:00 6.9 7.1 6.85 7.1 163821500
2012-06-05 00:00:00+00:00 7.1 7.14 6.9 6.91 147346700
2012-06-06 00:00:00+00:00 7.64 7.77 7.18 7.24 356852900
2012-06-07 00:00:00+00:00 7.42 7.9 7.38 7.81 276170400
2012-06-08 00:00:00+00:00 7.56 7.58 7.2 7.35 234555000
2012-06-11 00:00:00+00:00 7.28 7.77 7.28 7.72 204590700
2012-06-12 00:00:00+00:00 7.49 7.5 7.22 7.32 148255600
2012-06-13 00:00:00+00:00 7.5 7.61 7.35 7.4 162374500
2012-06-14 00:00:00+00:00 7.66 7.69 7.446 7.52 157528300
2012-06-15 00:00:00+00:00 7.9 7.9 7.55 7.72 211751900
2012-06-18 00:00:00+00:00 7.76 7.91 7.73 7.77 140319900
2012-06-19 00:00:00+00:00 8.109 8.207 7.86 7.88 248771500
2012-06-20 00:00:00+00:00 8.14 8.22 8.02 8.16 213582800
2012-06-21 00:00:00+00:00 7.82 8.17 7.8 8.109 227157600
2012-06-22 00:00:00+00:00 7.94 8.01 7.79 7.965 142219300
2012-06-25 00:00:00+00:00 7.6 7.74 7.56 7.735 151907200
2012-06-26 00:00:00+00:00 7.615 7.68 7.48 7.65 129776500
2012-06-27 00:00:00+00:00 7.77 7.82 7.61 7.68 111335500
2012-06-28 00:00:00+00:00 7.74 7.75 7.53 7.62 133125300
2012-06-29 00:00:00+00:00 8.18 8.2 7.95 8 258769800
2012-07-02 00:00:00+00:00 8.05 8.21 7.87 8.2 152974000
2012-07-03 00:00:00+00:00 8.06 8.119 8.01 8.06 57653700
2012-07-05 00:00:00+00:00 7.82 8.06 7.82 8.029 120162000
2012-07-06 00:00:00+00:00 7.66 7.8 7.65 7.69 116595200
2012-07-09 00:00:00+00:00 7.56 7.7 7.53 7.63 87078300
2012-07-10 00:00:00+00:00 7.48 7.67 7.4 7.65 101758300
2012-07-11 00:00:00+00:00 7.63 7.69 7.45 7.49 128619400
2012-07-12 00:00:00+00:00 7.48 7.55 7.43 7.53 107637000
2012-07-13 00:00:00+00:00 7.82 7.83 7.55 7.56 177154500
2012-07-16 00:00:00+00:00 7.81 7.96 7.77 7.925 109439900
2012-07-17 00:00:00+00:00 7.92 7.93 7.75 7.92 127266800
2012-07-18 00:00:00+00:00 7.53 7.93 7.5 7.88 254341100
2012-07-19 00:00:00+00:00 7.26 7.65 7.12 7.56 260959200
2012-07-20 00:00:00+00:00 7.07 7.21 7.06 7.205 160463000
2012-07-23 00:00:00+00:00 7.09 7.15 6.9 6.94 168638000
2012-07-24 00:00:00+00:00 7.04 7.18 6.97 7.16 138286700
2012-07-25 00:00:00+00:00 7.07 7.161 7.01 7.11 117241300
2012-07-26 00:00:00+00:00 7.17 7.24 7.11 7.22 121121500
2012-07-27 00:00:00+00:00 7.31 7.4 7.14 7.2 146210900
2012-07-30 00:00:00+00:00 7.28 7.38 7.26 7.28 74683800
2012-07-31 00:00:00+00:00 7.34 7.34 7.21 7.28 86093200
2012-08-01 00:00:00+00:00 7.22 7.36 7.21 7.35 97187800
2012-08-02 00:00:00+00:00 7.18 7.26 7.1 7.12 112848000
2012-08-03 00:00:00+00:00 7.43 7.49 7.27 7.3 130250800
2012-08-06 00:00:00+00:00 7.64 7.68 7.44 7.45 112657000
2012-08-07 00:00:00+00:00 7.67 7.85 7.66 7.71 119572200
2012-08-08 00:00:00+00:00 7.67 7.77 7.57 7.58 73353500
2012-08-09 00:00:00+00:00 7.72 7.76 7.67 7.68 59676500
2012-08-10 00:00:00+00:00 7.74 7.76 7.63 7.66 50443800
2012-08-13 00:00:00+00:00 7.72 7.83 7.68 7.72 58341300
2012-08-14 00:00:00+00:00 7.78 7.9 7.74 7.78 104644000
2012-08-15 00:00:00+00:00 7.87 7.87 7.73 7.75 73559200
2012-08-16 00:00:00+00:00 7.93 7.96 7.83 7.89 78181200
2012-08-17 00:00:00+00:00 8 8.119 7.94 7.97 138307600
2012-08-20 00:00:00+00:00 8.15 8.189 7.98 7.98 100276900
2012-08-21 00:00:00+00:00 8.189 8.4 8.16 8.25 190820700
2012-08-22 00:00:00+00:00 8.22 8.32 8.119 8.14 140333800
2012-08-23 00:00:00+00:00 8.15 8.27 8.119 8.22 98866200
2012-08-24 00:00:00+00:00 8.16 8.21 8.029 8.119 88822200
2012-08-27 00:00:00+00:00 8.07 8.2 8.05 8.2 96142100
2012-08-28 00:00:00+00:00 7.96 8.109 7.95 8.039 96280400
2012-08-29 00:00:00+00:00 8 8.1 7.98 7.98 106910900
2012-08-30 00:00:00+00:00 7.91 7.95 7.83 7.95 91900100
2012-08-31 00:00:00+00:00 7.99 8.02 7.88 8 91730500
2012-09-04 00:00:00+00:00 8 8.109 7.95 8 81461800
2012-09-05 00:00:00+00:00 7.95 8.02 7.93 7.99 55769500
2012-09-06 00:00:00+00:00 8.35 8.35 8.029 8.035 200570300
2012-09-07 00:00:00+00:00 8.8 8.8 8.45 8.465 232843000
2012-09-10 00:00:00+00:00 8.58 8.92 8.53 8.835 190045600
2012-09-11 00:00:00+00:00 9.03 9.05 8.58 8.63 201630400
2012-09-12 00:00:00+00:00 8.97 9.19 8.87 9.15 204167100
2012-09-13 00:00:00+00:00 9.4 9.48 8.81 8.89 330759200
2012-09-14 00:00:00+00:00 9.55 9.79 9.45 9.6 329619800
2012-09-17 00:00:00+00:00 9.3 9.49 9.27 9.39 141409300
2012-09-18 00:00:00+00:00 9.23 9.29 9.09 9.16 150503600
2012-09-19 00:00:00+00:00 9.29 9.46 9.26 9.33 126360100
2012-09-20 00:00:00+00:00 9.19 9.25 9.08 9.14 111354300
2012-09-21 00:00:00+00:00 9.11 9.35 9.08 9.35 155865300
2012-09-24 00:00:00+00:00 9.1 9.2 8.95 8.99 113331700
2012-09-25 00:00:00+00:00 8.925 9.21 8.91 9.165 146678000
2012-09-26 00:00:00+00:00 8.815 8.91 8.7 8.855 158640100
2012-09-27 00:00:00+00:00 8.97 9.07 8.899 8.935 119023800
2012-09-28 00:00:00+00:00 8.83 8.93 8.79 8.89 119039200
2012-10-01 00:00:00+00:00 8.96 9.13 8.85 8.87 133194500
2012-10-02 00:00:00+00:00 8.93 9.105 8.88 9.1 109601600
2012-10-03 00:00:00+00:00 9.11 9.12 8.91 8.975 115670100
2012-10-04 00:00:00+00:00 9.41 9.42 9.15 9.2 160486300
2012-10-05 00:00:00+00:00 9.32 9.65 9.23 9.56 204255800
2012-10-08 00:00:00+00:00 9.28 9.38 9.13 9.16 81944200
2012-10-09 00:00:00+00:00 9.21 9.38 9.04 9.325 153036100
2012-10-10 00:00:00+00:00 9.21 9.27 9.06 9.19 110640900
2012-10-11 00:00:00+00:00 9.34 9.42 9.3 9.39 122925800
2012-10-12 00:00:00+00:00 9.12 9.28 9.05 9.16 158562500
2012-10-15 00:00:00+00:00 9.44 9.44 9.16 9.235 154458500
2012-10-16 00:00:00+00:00 9.46 9.6 9.38 9.53 172663600
2012-10-17 00:00:00+00:00 9.44 9.6 9.3 9.42 229733500
2012-10-18 00:00:00+00:00 9.47 9.57 9.37 9.38 149739000
2012-10-19 00:00:00+00:00 9.44 9.55 9.39 9.42 169451300
2012-10-22 00:00:00+00:00 9.55 9.56 9.41 9.47 132617600
2012-10-23 00:00:00+00:00 9.36 9.47 9.28 9.44 160627600
2012-10-24 00:00:00+00:00 9.31 9.49 9.3 9.45 120868300
2012-10-25 00:00:00+00:00 9.24 9.415 9.17 9.365 121778900
2012-10-26 00:00:00+00:00 9.12 9.27 9.047 9.165 124687200
2012-10-31 00:00:00+00:00 9.32 9.35 9.15 9.2 94925500
2012-11-01 00:00:00+00:00 9.74 9.75 9.272 9.34 205700000
2012-11-02 00:00:00+00:00 9.85 9.97 9.77 9.87 220993300
2012-11-05 00:00:00+00:00 9.75 9.925 9.62 9.83 121104400
2012-11-06 00:00:00+00:00 9.94 9.97 9.75 9.825 132533500
2012-11-07 00:00:00+00:00 9.23 9.67 9.22 9.63 286124300
2012-11-08 00:00:00+00:00 9.39 9.6 9.38 9.49 223927000
2012-11-09 00:00:00+00:00 9.43 9.59 9.27 9.295 141093100
2012-11-12 00:00:00+00:00 9.39 9.52 9.38 9.5 68498300
2012-11-13 00:00:00+00:00 9.33 9.55 9.28 9.31 119629700
2012-11-14 00:00:00+00:00 8.99 9.42 8.95 9.38 197392900
2012-11-15 00:00:00+00:00 9.09 9.2 9.02 9.03 144569800
2012-11-16 00:00:00+00:00 9.12 9.21 8.92 9.12 178995000
2012-11-19 00:00:00+00:00 9.49 9.533 9.32 9.36 146244000
2012-11-20 00:00:00+00:00 9.63 9.68 9.42 9.465 150696200
2012-11-21 00:00:00+00:00 9.77 9.78 9.63 9.67 132499300
2012-11-23 00:00:00+00:00 9.9 9.9 9.8 9.835 59048000
2012-11-26 00:00:00+00:00 9.835 9.87 9.75 9.82 99730900
2012-11-27 00:00:00+00:00 9.66 9.95 9.66 9.89 149684000
2012-11-28 00:00:00+00:00 9.76 9.76 9.38 9.56 152946500
2012-11-29 00:00:00+00:00 9.83 9.88 9.76 9.84 125905400
2012-11-30 00:00:00+00:00 9.86 9.87 9.755 9.78 108739700
2012-12-03 00:00:00+00:00 9.8 9.94 9.78 9.93 99872800
2012-12-04 00:00:00+00:00 9.91 9.91 9.77 9.79 144124500
2012-12-05 00:00:00+00:00 10.46 10.56 9.95 9.97 463491000
2012-12-06 00:00:00+00:00 10.455 10.58 10.29 10.51 176607900
2012-12-07 00:00:00+00:00 10.635 10.68 10.48 10.56 192055100
2012-12-10 00:00:00+00:00 10.57 10.63 10.46 10.6 148065200
2012-12-11 00:00:00+00:00 10.51 10.71 10.5 10.64 159121300
2012-12-12 00:00:00+00:00 10.61 10.7 10.51 10.59 166819300
2012-12-13 00:00:00+00:00 10.54 10.66 10.51 10.59 106232100
2012-12-14 00:00:00+00:00 10.58 10.6 10.52 10.545 91707200
2012-12-17 00:00:00+00:00 11 11 10.64 10.645 170087700
2012-12-18 00:00:00+00:00 11.36 11.36 11.05 11.24 255238400
2012-12-19 00:00:00+00:00 11.19 11.49 11.17 11.4 193013700
2012-12-20 00:00:00+00:00 11.52 11.52 11.08 11.1 184449000
2012-12-21 00:00:00+00:00 11.29 11.35 11.12 11.2 244892300
2012-12-24 00:00:00+00:00 11.25 11.29 11.21 11.27 50657100
2012-12-26 00:00:00+00:00 11.54 11.63 11.27 11.29 146097900
2012-12-27 00:00:00+00:00 11.47 11.69 11.23 11.66 210411400
2012-12-28 00:00:00+00:00 11.36 11.49 11.27 11.32 131872200
2012-12-31 00:00:00+00:00 11.61 11.65 11.3 11.37 170837500
2013-01-02 00:00:00+00:00 12.03 12.15 11.9 12.05 236021400
2013-01-03 00:00:00+00:00 11.96 12.05 11.88 12.01 157149700
2013-01-04 00:00:00+00:00 12.11 12.11 11.93 11.97 132601900
2013-01-07 00:00:00+00:00 12.09 12.2 12 12.15 201403500
2013-01-08 00:00:00+00:00 11.98 12.1 11.89 12.09 168461100
2013-01-09 00:00:00+00:00 11.43 12 11.33 11.87 335692000
2013-01-10 00:00:00+00:00 11.78 11.81 11.54 11.61 199964900
2013-01-11 00:00:00+00:00 11.63 11.72 11.51 11.7 146136700
2013-01-14 00:00:00+00:00 11.47 11.63 11.38 11.61 110032900
2013-01-15 00:00:00+00:00 11.55 11.62 11.3 11.38 126058300
2013-01-16 00:00:00+00:00 11.78 11.79 11.47 11.58 164579300
2013-01-17 00:00:00+00:00 11.28 11.7 11.17 11.69 323622200
2013-01-18 00:00:00+00:00 11.14 11.33 11.02 11.26 179971800
2013-01-22 00:00:00+00:00 11.35 11.36 11.09 11.12 137242700
2013-01-23 00:00:00+00:00 11.42 11.44 11.23 11.38 121638200
2013-01-24 00:00:00+00:00 11.53 11.67 11.42 11.45 127870800
2013-01-25 00:00:00+00:00 11.62 11.72 11.51 11.69 100171600
2013-01-28 00:00:00+00:00 11.48 11.67 11.43 11.64 91304200
2013-01-29 00:00:00+00:00 11.49 11.58 11.4 11.42 96824100
2013-01-30 00:00:00+00:00 11.38 11.54 11.35 11.49 89130200
2013-01-31 00:00:00+00:00 11.32 11.36 11.22 11.32 97024700
2013-02-01 00:00:00+00:00 11.71 11.73 11.37 11.41 161141300
2013-02-04 00:00:00+00:00 11.48 11.69 11.46 11.58 139733000
2013-02-05 00:00:00+00:00 11.88 11.98 11.56 11.59 187785900
2013-02-06 00:00:00+00:00 11.93 11.97 11.73 11.73 173626800
2013-02-07 00:00:00+00:00 11.84 11.98 11.73 11.97 173209900
2013-02-08 00:00:00+00:00 11.76 11.9 11.72 11.86 145321100
2013-02-11 00:00:00+00:00 11.86 11.9 11.67 11.73 103510800
2013-02-12 00:00:00+00:00 12.245 12.34 11.78 11.87 232173900
2013-02-13 00:00:00+00:00 12.17 12.42 12.05 12.35 192742500
2013-02-14 00:00:00+00:00 12.13 12.27 12.07 12.09 144104800
2013-02-15 00:00:00+00:00 12.03 12.21 11.97 12.205 158242700
2013-02-19 00:00:00+00:00 12.19 12.31 12.06 12.1 170783400
2013-02-20 00:00:00+00:00 11.8 12.29 11.75 12.18 193466400
2013-02-21 00:00:00+00:00 11.42 11.73 11.35 11.725 235532400
2013-02-22 00:00:00+00:00 11.44 11.635 11.26 11.62 179268200
2013-02-25 00:00:00+00:00 11.03 11.61 10.98 11.6 206893900
2013-02-26 00:00:00+00:00 11.13 11.22 11.03 11.12 173039600
2013-02-27 00:00:00+00:00 11.3 11.36 11.1 11.15 147145400
2013-02-28 00:00:00+00:00 11.23 11.37 11.195 11.325 143546700
2013-03-01 00:00:00+00:00 11.34 11.55 11.02 11.13 189050700
2013-03-04 00:00:00+00:00 11.41 11.45 11.22 11.27 116436000
2013-03-05 00:00:00+00:00 11.55 11.709 11.53 11.56 135869700
2013-03-06 00:00:00+00:00 11.92 12.02 11.76 11.77 182557500
2013-03-07 00:00:00+00:00 12.26 12.28 11.98 12 212403800
2013-03-08 00:00:00+00:00 12.07 12.44 12.02 12.42 209777100
2013-03-11 00:00:00+00:00 12.15 12.22 12.02 12.08 106427800
2013-03-12 00:00:00+00:00 12.01 12.18 11.91 12.11 128134700
2013-03-13 00:00:00+00:00 12.06 12.11 11.98 12.04 86216800
2013-03-14 00:00:00+00:00 12.11 12.19 12.1 12.12 115440000
2013-03-15 00:00:00+00:00 12.57 12.66 12.35 12.52 319019100
2013-03-18 00:00:00+00:00 12.56 12.68 12.26 12.29 189783300
2013-03-19 00:00:00+00:00 12.71 12.94 12.59 12.79 242939600
2013-03-20 00:00:00+00:00 12.78 12.89 12.71 12.785 219121700
2013-03-21 00:00:00+00:00 12.57 12.84 12.55 12.71 154529400
2013-03-22 00:00:00+00:00 12.56 12.67 12.48 12.62 101974400
2013-03-25 00:00:00+00:00 12.4 12.72 12.32 12.68 154320200
2013-03-26 00:00:00+00:00 12.28 12.5 12.15 12.45 135654700
2013-03-27 00:00:00+00:00 12.23 12.28 12.12 12.14 107177200
2013-03-28 00:00:00+00:00 12.18 12.28 12.11 12.24 92013900
2013-04-01 00:00:00+00:00 12.15 12.28 12.1 12.15 86281600
2013-04-02 00:00:00+00:00 12.15 12.245 12.14 12.24 102626000
2013-04-03 00:00:00+00:00 11.81 12.14 11.72 12.115 199765800
2013-04-04 00:00:00+00:00 11.94 11.99 11.72 11.81 117831400
2013-04-05 00:00:00+00:00 11.97 12.01 11.64 11.67 141061900
2013-04-08 00:00:00+00:00 12.21 12.21 11.91 12 101419200
2013-04-09 00:00:00+00:00 12.25 12.35 12.21 12.25 132365800
2013-04-10 00:00:00+00:00 12.32 12.4 12.26 12.31 105853900
2013-04-11 00:00:00+00:00 12.27 12.33 12.16 12.31 100241900
2013-04-12 00:00:00+00:00 12.17 12.25 12.07 12.145 88191700
2013-04-15 00:00:00+00:00 11.98 12.32 11.97 12.19 176504300
2013-04-16 00:00:00+00:00 12.28 12.36 12.08 12.21 147641900
2013-04-17 00:00:00+00:00 11.7 12.02 11.45 11.91 335408600
2013-04-18 00:00:00+00:00 11.44 11.65 11.23 11.61 220290500
2013-04-19 00:00:00+00:00 11.66 11.69 11.43 11.56 119792000
2013-04-22 00:00:00+00:00 11.72 11.75 11.57 11.68 88532300
2013-04-23 00:00:00+00:00 12.07 12.16 11.9 11.92 176816000
2013-04-24 00:00:00+00:00 12.31 12.37 12.12 12.135 120624900
2013-04-25 00:00:00+00:00 12.44 12.54 12.36 12.39 118694600
2013-04-26 00:00:00+00:00 12.42 12.46 12.29 12.34 83093200
2013-04-29 00:00:00+00:00 12.38 12.48 12.37 12.45 65432400
2013-04-30 00:00:00+00:00 12.31 12.4 12.2 12.39 89362100
2013-05-01 00:00:00+00:00 12.14 12.27 12.08 12.2 88417800
2013-05-02 00:00:00+00:00 12.19 12.21 12.05 12.17 76530500
2013-05-03 00:00:00+00:00 12.24 12.41 12.205 12.36 94322600
2013-05-06 00:00:00+00:00 12.88 12.89 12.36 12.39 265062800
2013-05-07 00:00:00+00:00 12.9 13.11 12.76 12.92 218248100
2013-05-08 00:00:00+00:00 13.02 13.18 12.84 12.85 149755400
2013-05-09 00:00:00+00:00 12.91 13.06 12.87 13.05 112952300
2013-05-10 00:00:00+00:00 13.02 13.04 12.89 12.94 87625100
2013-05-13 00:00:00+00:00 12.98 13.1 12.95 12.98 94295600
2013-05-14 00:00:00+00:00 13.34 13.36 13.02 13.04 154128100
2013-05-15 00:00:00+00:00 13.44 13.55 13.29 13.29 139728300
2013-05-16 00:00:00+00:00 13.36 13.55 13.32 13.41 117977000
2013-05-17 00:00:00+00:00 13.43 13.52 13.39 13.5 107335600
2013-05-20 00:00:00+00:00 13.51 13.6 13.39 13.39 88941700
2013-05-21 00:00:00+00:00 13.44 13.56 13.36 13.525 111810300
2013-05-22 00:00:00+00:00 13.31 13.73 13.17 13.485 174531500
2013-05-23 00:00:00+00:00 13.21 13.42 12.82 12.93 190522000
2013-05-24 00:00:00+00:00 13.24 13.26 13.12 13.17 83452600
2013-05-28 00:00:00+00:00 13.35 13.51 13.31 13.49 133553000
2013-05-29 00:00:00+00:00 13.48 13.555 13.27 13.31 126397300
2013-05-30 00:00:00+00:00 13.83 13.93 13.43 13.5 153383400
2013-05-31 00:00:00+00:00 13.66 13.99 13.652 13.91 160176100
2013-06-03 00:00:00+00:00 13.55 13.73 13.21 13.69 197168200
2013-06-04 00:00:00+00:00 13.36 13.67 13.31 13.51 130055400
2013-06-05 00:00:00+00:00 13.09 13.45 12.97 13.285 185778900
2013-06-06 00:00:00+00:00 13.2 13.24 12.85 13.075 142361900
2013-06-07 00:00:00+00:00 13.38 13.39 13.15 13.33 121057300
2013-06-10 00:00:00+00:00 13.3 13.55 13.26 13.49 104648500
2013-06-11 00:00:00+00:00 13.12 13.28 13.07 13.11 106369100
2013-06-12 00:00:00+00:00 13.06 13.25 12.97 13.24 117289400
2013-06-13 00:00:00+00:00 13.21 13.26 12.97 13 103636400
2013-06-14 00:00:00+00:00 13.07 13.24 13.03 13.23 101380700
2013-06-17 00:00:00+00:00 13.21 13.262 13.13 13.17 115441500
2013-06-18 00:00:00+00:00 13.27 13.34 13.2 13.215 67257600
2013-06-19 00:00:00+00:00 13.19 13.4 13.17 13.285 103850000
2013-06-20 00:00:00+00:00 12.89 13.14 12.8 13.08 185741100
2013-06-21 00:00:00+00:00 12.69 13 12.39 12.99 196004700
2013-06-24 00:00:00+00:00 12.3 12.45 12.13 12.4 159762500
2013-06-25 00:00:00+00:00 12.67 12.77 12.43 12.61 132844500
2013-06-26 00:00:00+00:00 12.76 12.853 12.68 12.84 105282000
2013-06-27 00:00:00+00:00 13.01 13.03 12.78 12.84 124616300
2013-06-28 00:00:00+00:00 12.86 13 12.84 12.97 90297100
2013-07-01 00:00:00+00:00 12.93 13.1 12.92 12.95 83146300
2013-07-02 00:00:00+00:00 12.9 13.1 12.8 12.95 83708600
2013-07-03 00:00:00+00:00 12.83 12.84 12.73 12.82 37971700
2013-07-05 00:00:00+00:00 13.06 13.08 12.91 12.99 80758500
2013-07-08 00:00:00+00:00 13.28 13.37 13.08 13.11 107441200
2013-07-09 00:00:00+00:00 13.53 13.53 13.25 13.37 106067300
2013-07-10 00:00:00+00:00 13.37 13.53 13.31 13.51 104353400
2013-07-11 00:00:00+00:00 13.51 13.58 13.33 13.54 93756500
2013-07-12 00:00:00+00:00 13.78 13.8 13.47 13.54 123958400
2013-07-15 00:00:00+00:00 13.88 13.94 13.76 13.93 90765200
2013-07-16 00:00:00+00:00 13.92 14.02 13.77 13.93 146041900
2013-07-17 00:00:00+00:00 14.31 14.44 14.04 14.055 249674500
2013-07-18 00:00:00+00:00 14.76 14.85 14.4 14.4 221796400
2013-07-19 00:00:00+00:00 14.75 14.76 14.6 14.76 136196600
2013-07-22 00:00:00+00:00 14.92 14.99 14.65 14.74 112416800
2013-07-23 00:00:00+00:00 14.94 15.03 14.86 14.98 113736000
2013-07-24 00:00:00+00:00 14.71 15 14.68 15 117771100
2013-07-25 00:00:00+00:00 14.83 14.85 14.54 14.62 89243300
2013-07-26 00:00:00+00:00 14.73 14.76 14.62 14.7 73777800
2013-07-29 00:00:00+00:00 14.52 14.7 14.45 14.65 88616700
2013-07-30 00:00:00+00:00 14.52 14.63 14.46 14.58 71675200
2013-07-31 00:00:00+00:00 14.6 14.85 14.57 14.58 128182500
2013-08-01 00:00:00+00:00 14.95 14.97 14.82 14.845 108554200
2013-08-02 00:00:00+00:00 14.84 14.91 14.78 14.88 83367400
2013-08-05 00:00:00+00:00 14.8 14.84 14.7 14.77 61711900
2013-08-06 00:00:00+00:00 14.64 14.76 14.63 14.74 84249800
2013-08-07 00:00:00+00:00 14.53 14.57 14.25 14.43 128693600
2013-08-08 00:00:00+00:00 14.61 14.7 14.44 14.675 88544100
2013-08-09 00:00:00+00:00 14.45 14.61 14.4 14.56 72897900
2013-08-12 00:00:00+00:00 14.41 14.47 14.32 14.355 60867500
2013-08-13 00:00:00+00:00 14.51 14.62 14.33 14.47 75777900
2013-08-14 00:00:00+00:00 14.6 14.75 14.54 14.55 78553700
2013-08-15 00:00:00+00:00 14.32 14.47 14.31 14.46 99238900
2013-08-16 00:00:00+00:00 14.42 14.47 14.21 14.29 104959000
2013-08-19 00:00:00+00:00 14.15 14.38 14.13 14.37 98384600
2013-08-20 00:00:00+00:00 14.29 14.38 13.98 14.18 114332500
2013-08-21 00:00:00+00:00 14.34 14.48 14.2 14.26 91680600
2013-08-22 00:00:00+00:00 14.57 14.6 14.4 14.47 76348300
2013-08-23 00:00:00+00:00 14.57 14.69 14.51 14.59 67699200
2013-08-26 00:00:00+00:00 14.49 14.7 14.49 14.53 68938900
2013-08-27 00:00:00+00:00 14.11 14.41 14.1 14.26 122534200
2013-08-28 00:00:00+00:00 14.12 14.25 13.99 14.07 96473600
2013-08-29 00:00:00+00:00 14.17 14.3 14.12 14.14 76691500
2013-08-30 00:00:00+00:00 14.12 14.25 14.07 14.23 73233400
2013-09-03 00:00:00+00:00 14.25 14.38 14.18 14.3 74628700
2013-09-04 00:00:00+00:00 14.32 14.4 14.205 14.215 71132500
2013-09-05 00:00:00+00:00 14.37 14.55 14.36 14.41 71229200
2013-09-06 00:00:00+00:00 14.36 14.5 14.27 14.47 75513100
2013-09-09 00:00:00+00:00 14.48 14.49 14.37 14.41 53447300
2013-09-10 00:00:00+00:00 14.61 14.69 14.52 14.52 93180300
2013-09-11 00:00:00+00:00 14.65 14.68 14.52 14.56 81098900
2013-09-12 00:00:00+00:00 14.48 14.68 14.46 14.65 63355900
2013-09-13 00:00:00+00:00 14.49 14.55 14.45 14.46 56793400
2013-09-16 00:00:00+00:00 14.53 14.68 14.53 14.61 65276500
2013-09-17 00:00:00+00:00 14.55 14.62 14.4 14.53 73686100
2013-09-18 00:00:00+00:00 14.715 14.82 14.49 14.53 127092600
2013-09-19 00:00:00+00:00 14.61 14.83 14.58 14.8 79710700
2013-09-20 00:00:00+00:00 14.44 14.66 14.44 14.655 145174800
2013-09-23 00:00:00+00:00 14.14 14.325 14.09 14.31 127283100
2013-09-24 00:00:00+00:00 14.09 14.25 14.01 14.18 96912800
2013-09-25 00:00:00+00:00 14.14 14.24 14.03 14.12 86546100
2013-09-26 00:00:00+00:00 14.08 14.22 13.98 14.19 89352900
2013-09-27 00:00:00+00:00 13.9 14.03 13.88 13.99 79156000
2013-09-30 00:00:00+00:00 13.8 13.93 13.6 13.685 104439900
2013-10-01 00:00:00+00:00 13.9 13.92 13.81 13.85 57631000
2013-10-02 00:00:00+00:00 14.06 14.1 13.79 13.81 89186300
2013-10-03 00:00:00+00:00 14 14.1 13.83 14.07 95269000
2013-10-04 00:00:00+00:00 14.05 14.08 13.96 14.02 63495400
2013-10-07 00:00:00+00:00 13.81 13.95 13.8 13.91 64822700
2013-10-08 00:00:00+00:00 13.69 13.9 13.68 13.83 92746200
2013-10-09 00:00:00+00:00 13.84 13.93 13.69 13.72 96012800
2013-10-10 00:00:00+00:00 14.23 14.25 14.01 14.03 102305800
2013-10-11 00:00:00+00:00 14.19 14.28 14.11 14.24 83422000
2013-10-14 00:00:00+00:00 14.35 14.39 14.05 14.06 78367500
2013-10-15 00:00:00+00:00 14.24 14.43 14.21 14.39 102099300
2013-10-16 00:00:00+00:00 14.56 14.61 14.315 14.32 153481000
2013-10-17 00:00:00+00:00 14.66 14.66 14.41 14.445 94090800
2013-10-18 00:00:00+00:00 14.63 14.72 14.53 14.68 96986400
2013-10-21 00:00:00+00:00 14.52 14.59 14.47 14.53 91280600
2013-10-22 00:00:00+00:00 14.52 14.63 14.45 14.55 102836000
2013-10-23 00:00:00+00:00 14.21 14.48 14.17 14.45 108529700
2013-10-24 00:00:00+00:00 14.17 14.2 14.07 14.12 92566100
2013-10-25 00:00:00+00:00 14.26 14.27 14.14 14.15 58592400
2013-10-28 00:00:00+00:00 14.23 14.32 14.21 14.29 62328000
2013-10-29 00:00:00+00:00 14.15 14.29 14.06 14.26 79563100
2013-10-30 00:00:00+00:00 14.17 14.29 14.15 14.19 76624200
2013-10-31 00:00:00+00:00 13.97 14.16 13.96 14.135 105713400
2013-11-01 00:00:00+00:00 14.02 14.1 13.96 13.995 75610400
2013-11-04 00:00:00+00:00 14.04 14.12 14 14.09 53063200
2013-11-05 00:00:00+00:00 13.93 14.02 13.9 13.99 71809700
2013-11-06 00:00:00+00:00 13.96 14.02 13.91 13.99 66712900
2013-11-07 00:00:00+00:00 13.8 14.03 13.8 14 80775000
2013-11-08 00:00:00+00:00 14.32 14.32 13.835 13.86 158227300
2013-11-11 00:00:00+00:00 14.4 14.42 14.24 14.27 66357400
2013-11-12 00:00:00+00:00 14.32 14.46 14.26 14.33 69498300
2013-11-13 00:00:00+00:00 14.64 14.64 14.2 14.255 128018500
2013-11-14 00:00:00+00:00 14.795 14.82 14.61 14.675 128999400
2013-11-15 00:00:00+00:00 14.92 14.95 14.77 14.79 105475300
2013-11-18 00:00:00+00:00 14.92 15.17 14.86 14.975 145554400
2013-11-19 00:00:00+00:00 15.2 15.3 14.97 14.975 161505600
2013-11-20 00:00:00+00:00 15.14 15.27 15.08 15.25 107174900
2013-11-21 00:00:00+00:00 15.59 15.6 15.2 15.22 142053600
2013-11-22 00:00:00+00:00 15.64 15.787 15.6 15.695 110331100
2013-11-25 00:00:00+00:00 15.81 15.98 15.66 15.67 153765500
2013-11-26 00:00:00+00:00 15.88 15.97 15.81 15.835 114833700
2013-11-27 00:00:00+00:00 15.83 15.94 15.76 15.87 75503600
2013-11-29 00:00:00+00:00 15.82 15.92 15.79 15.835 44288400
2013-12-02 00:00:00+00:00 15.73 15.97 15.7 15.84 92987400
2013-12-03 00:00:00+00:00 15.54 15.79 15.39 15.64 106450300
2013-12-04 00:00:00+00:00 15.63 15.73 15.36 15.39 96678100
2013-12-05 00:00:00+00:00 15.43 15.64 15.36 15.61 95230500
2013-12-06 00:00:00+00:00 15.56 15.72 15.46 15.61 92203100
2013-12-09 00:00:00+00:00 15.58 15.67 15.56 15.62 50167200
2013-12-10 00:00:00+00:00 15.56 15.675 15.51 15.52 67486300
2013-12-11 00:00:00+00:00 15.25 15.55 15.18 15.55 117565000
2013-12-12 00:00:00+00:00 15.25 15.35 15.12 15.26 76113900
2013-12-13 00:00:00+00:00 15.18 15.33 15.13 15.275 61575600
2013-12-16 00:00:00+00:00 15.24 15.34 15.2 15.22 71467000
2013-12-17 00:00:00+00:00 15.18 15.295 15.14 15.24 74247300
2013-12-18 00:00:00+00:00 15.69 15.71 15.06 15.28 152929300
2013-12-19 00:00:00+00:00 15.75 15.79 15.6 15.66 97776100
2013-12-20 00:00:00+00:00 15.6 15.87 15.6 15.75 131956000
2013-12-23 00:00:00+00:00 15.69 15.79 15.69 15.72 52565800
2013-12-24 00:00:00+00:00 15.7 15.75 15.67 15.72 21770400
2013-12-26 00:00:00+00:00 15.65 15.75 15.64 15.735 48803600
2013-12-27 00:00:00+00:00 15.67 15.7 15.58 15.67 40049300
2013-12-30 00:00:00+00:00 15.54 15.69 15.52 15.64 56218800
2013-12-31 00:00:00+00:00 15.57 15.615 15.51 15.6 57188900
2014-01-02 00:00:00+00:00 16.1 16.16 15.68 15.69 148709900
2014-01-03 00:00:00+00:00 16.41 16.5 16.23 16.27 129921800
2014-01-06 00:00:00+00:00 16.66 16.73 16.56 16.625 114431300
2014-01-07 00:00:00+00:00 16.5 16.79 16.45 16.77 110605100
2014-01-08 00:00:00+00:00 16.58 16.69 16.519 16.67 101036400
2014-01-09 00:00:00+00:00 16.83 16.926 16.615 16.67 100947200
2014-01-10 00:00:00+00:00 16.77 16.79 16.61 16.75 87454100
2014-01-13 00:00:00+00:00 16.43 16.8 16.4 16.79 90025400
2014-01-14 00:00:00+00:00 16.77 16.77 16.53 16.54 97786600
2014-01-15 00:00:00+00:00 17.15 17.42 17.11 17.23 329333100
2014-01-16 00:00:00+00:00 17.08 17.14 16.99 17.09 163613100
2014-01-17 00:00:00+00:00 17.01 17.22 16.99 17.2 96232200
2014-01-21 00:00:00+00:00 17.01 17.155 16.87 17.08 118269200
2014-01-22 00:00:00+00:00 17.15 17.15 17 17.08 68114400
2014-01-23 00:00:00+00:00 16.86 17.09 16.74 17.07 123765900
2014-01-24 00:00:00+00:00 16.45 16.72 16.45 16.67 112899300
2014-01-27 00:00:00+00:00 16.309 16.54 16.059 16.37 134872600
2014-01-28 00:00:00+00:00 16.73 16.76 16.37 16.45 92891300
2014-01-29 00:00:00+00:00 16.68 16.85 16.49 16.55 130026900
2014-01-30 00:00:00+00:00 16.93 16.99 16.765 16.81 91028200
2014-01-31 00:00:00+00:00 16.75 16.98 16.61 16.72 139432100
2014-02-03 00:00:00+00:00 16.35 16.88 16.3 16.77 159981200
2014-02-04 00:00:00+00:00 16.35 16.56 16.26 16.49 125337000
2014-02-05 00:00:00+00:00 16.4 16.45 16.149 16.309 110576200
2014-02-06 00:00:00+00:00 16.69 16.73 16.44 16.45 110029100
2014-02-07 00:00:00+00:00 16.82 16.88 16.63 16.82 122571100
2014-02-10 00:00:00+00:00 16.72 16.73 16.6 16.71 95792500
2014-02-11 00:00:00+00:00 16.88 16.92 16.635 16.725 92617500
2014-02-12 00:00:00+00:00 16.75 16.91 16.65 16.87 96948900
2014-02-13 00:00:00+00:00 16.75 16.82 16.63 16.645 95756900
2014-02-14 00:00:00+00:00 16.7 16.76 16.65 16.74 100684300
2014-02-18 00:00:00+00:00 16.47 16.74 16.47 16.7 121778600
2014-02-19 00:00:00+00:00 16.2 16.45 16.18 16.379 137734500
2014-02-20 00:00:00+00:00 16.3 16.36 16.18 16.219 107425900
2014-02-21 00:00:00+00:00 16.29 16.485 16.25 16.29 106236900
2014-02-24 00:00:00+00:00 16.53 16.63 16.3 16.3 102909600
2014-02-25 00:00:00+00:00 16.34 16.56 16.32 16.5 81457000
2014-02-26 00:00:00+00:00 16.329 16.4 16.129 16.36 89831600
2014-02-27 00:00:00+00:00 16.49 16.49 16.21 16.27 71314500
2014-02-28 00:00:00+00:00 16.53 16.65 16.344 16.49 126252100
2014-03-03 00:00:00+00:00 16.3 16.41 16.19 16.3 87526800
2014-03-04 00:00:00+00:00 16.73 16.74 16.43 16.48 100849900
2014-03-05 00:00:00+00:00 17.25 17.31 16.78 16.79 207180000
2014-03-06 00:00:00+00:00 17.35 17.631 17.31 17.42 138078900
2014-03-07 00:00:00+00:00 17.33 17.59 17.246 17.54 109547100
2014-03-10 00:00:00+00:00 17.47 17.47 17.25 17.27 81732700
2014-03-11 00:00:00+00:00 17.27 17.52 17.25 17.51 90183400
2014-03-12 00:00:00+00:00 17.28 17.33 17.07 17.18 83309200
2014-03-13 00:00:00+00:00 17.16 17.47 17.07 17.34 100875700
2014-03-14 00:00:00+00:00 16.8 17.22 16.76 17.08 131297200
2014-03-17 00:00:00+00:00 17.11 17.17 16.97 16.98 79942200
2014-03-18 00:00:00+00:00 17.19 17.22 17.07 17.14 67402400
2014-03-19 00:00:00+00:00 17.44 17.49 17.12 17.16 105006500
2014-03-20 00:00:00+00:00 17.92 18 17.43 17.44 167048700
2014-03-21 00:00:00+00:00 17.56 18.03 17.56 18.03 155721000
2014-03-24 00:00:00+00:00 17.37 17.65 17.3 17.62 118500000
2014-03-25 00:00:00+00:00 17.21 17.58 17.2 17.49 98348600
2014-03-26 00:00:00+00:00 17.18 17.4 17.18 17.38 97573700
2014-03-27 00:00:00+00:00 17.01 17.49 16.83 17.28 175139500
2014-03-28 00:00:00+00:00 16.98 17.15 16.85 17.085 79909500
2014-03-31 00:00:00+00:00 17.2 17.27 17.12 17.15 62115800
2014-04-01 00:00:00+00:00 17.34 17.4 17.26 17.28 57423800
2014-04-02 00:00:00+00:00 17.23 17.38 17.13 17.35 65349000
2014-04-03 00:00:00+00:00 17.15 17.24 17.03 17.21 56534500
2014-04-04 00:00:00+00:00 16.72 17.22 16.7 17.21 110409600
2014-04-07 00:00:00+00:00 16.379 16.7 16.19 16.69 129464900
2014-04-08 00:00:00+00:00 16.44 16.5 16.25 16.37 75769700
2014-04-09 00:00:00+00:00 16.62 16.63 16.344 16.55 83096400
2014-04-10 00:00:00+00:00 16.12 16.62 16.1 16.62 98390900
2014-04-11 00:00:00+00:00 15.77 16.129 15.62 15.86 133757100
2014-04-14 00:00:00+00:00 16 16.219 15.78 16.05 99249400
2014-04-15 00:00:00+00:00 16.39 16.41 15.96 16.09 134956600
2014-04-16 00:00:00+00:00 16.129 16.219 15.78 16.2 172948900
2014-04-17 00:00:00+00:00 16.149 16.25 15.93 16.14 104765500
2014-04-21 00:00:00+00:00 16.09 16.17 16.03 16.149 51280500
2014-04-22 00:00:00+00:00 16.29 16.34 16.04 16.09 77888900
2014-04-23 00:00:00+00:00 16.37 16.4 16.23 16.3 52425600
2014-04-24 00:00:00+00:00 16.34 16.5 16.21 16.43 73910900
2014-04-25 00:00:00+00:00 15.95 16.23 15.93 16.1 84617700
2014-04-28 00:00:00+00:00 14.95 15.41 14.86 15.33 344935200
2014-04-29 00:00:00+00:00 15.24 15.3 14.91 15.04 155495000
2014-04-30 00:00:00+00:00 15.14 15.27 15.13 15.26 82359700
2014-05-01 00:00:00+00:00 15.09 15.217 15.03 15.14 68368300
2014-05-02 00:00:00+00:00 15.25 15.29 15.1 15.18 73562900
2014-05-05 00:00:00+00:00 15.08 15.13 15.02 15.075 51669100
2014-05-06 00:00:00+00:00 14.73 15.03 14.72 15.03 96268300
2014-05-07 00:00:00+00:00 14.8 14.92 14.75 14.86 91733000
2014-05-08 00:00:00+00:00 14.93 15.04 14.8 14.85 65232700
2014-05-09 00:00:00+00:00 14.74 14.985 14.67 14.95 86751600
2014-05-12 00:00:00+00:00 15.07 15.1 14.79 14.79 70669900
2014-05-13 00:00:00+00:00 15.03 15.1 14.94 15.04 58469100
2014-05-14 00:00:00+00:00 14.84 15.02 14.81 14.99 52431900
2014-05-15 00:00:00+00:00 14.55 14.83 14.39 14.82 104193000
2014-05-16 00:00:00+00:00 14.51 14.54 14.367 14.51 80476100
2014-05-19 00:00:00+00:00 14.67 14.7 14.38 14.425 51034700
2014-05-20 00:00:00+00:00 14.53 14.69 14.48 14.65 60108200
2014-05-21 00:00:00+00:00 14.61 14.71 14.5 14.59 64808600
2014-05-22 00:00:00+00:00 14.71 14.72 14.55 14.58 51929400
2014-05-23 00:00:00+00:00 14.72 14.815 14.65 14.71 51648900
2014-05-27 00:00:00+00:00 15.22 15.31 14.98 15 125398900
2014-05-28 00:00:00+00:00 15.14 15.28 15.03 15.27 72481700
2014-05-29 00:00:00+00:00 15.15 15.21 15.1 15.12 39423000
2014-05-30 00:00:00+00:00 15.14 15.23 15.07 15.12 45787100
2014-06-02 00:00:00+00:00 15.26 15.28 15.05 15.16 46374000
2014-06-03 00:00:00+00:00 15.21 15.28 15.11 15.2 49025400
2014-06-04 00:00:00+00:00 15.21 15.26 15.15 15.19 39757300
2014-06-05 00:00:00+00:00 15.43 15.48 15.22 15.28 60300500
2014-06-06 00:00:00+00:00 15.59 15.65 15.43 15.45 73930500
2014-06-09 00:00:00+00:00 15.84 15.88 15.6 15.6 78647900
2014-06-10 00:00:00+00:00 15.92 15.94 15.73 15.78 49852900
2014-06-11 00:00:00+00:00 15.59 15.82 15.55 15.69 73729300
2014-06-12 00:00:00+00:00 15.42 15.66 15.38 15.57 66173200
2014-06-13 00:00:00+00:00 15.44 15.55 15.33 15.46 61588100
2014-06-16 00:00:00+00:00 15.28 15.31 15.18 15.3 55425400
2014-06-17 00:00:00+00:00 15.59 15.62 15.255 15.265 58011600
2014-06-18 00:00:00+00:00 15.65 15.69 15.44 15.63 70220300
2014-06-19 00:00:00+00:00 15.55 15.67 15.5 15.66 40953900
2014-06-20 00:00:00+00:00 15.45 15.63 15.44 15.63 54765300
2014-06-23 00:00:00+00:00 15.64 15.65 15.38 15.47 58492200
2014-06-24 00:00:00+00:00 15.49 15.7 15.44 15.555 73262500
2014-06-25 00:00:00+00:00 15.47 15.49 15.28 15.45 71766400
2014-06-26 00:00:00+00:00 15.41 15.53 15.27 15.44 68519600
2014-06-27 00:00:00+00:00 15.33 15.42 15.31 15.38 58149200
2014-06-30 00:00:00+00:00 15.37 15.45 15.29 15.31 47776800
2014-07-01 00:00:00+00:00 15.6 15.65 15.38 15.38 74193500
2014-07-02 00:00:00+00:00 15.85 16.03 15.76 15.78 87685800
2014-07-03 00:00:00+00:00 16.03 16.23 16 16.07 70582300
2014-07-07 00:00:00+00:00 15.94 16 15.83 15.99 62437100
2014-07-08 00:00:00+00:00 15.58 15.83 15.52 15.82 73343400
2014-07-09 00:00:00+00:00 15.6 15.69 15.54 15.62 46415800
2014-07-10 00:00:00+00:00 15.44 15.53 15.25 15.33 61458500
2014-07-11 00:00:00+00:00 15.38 15.43 15.3 15.39 56849400
2014-07-14 00:00:00+00:00 15.57 15.67 15.52 15.62 59181700
2014-07-15 00:00:00+00:00 15.81 15.85 15.66 15.75 99946700
2014-07-16 00:00:00+00:00 15.51 15.66 15.43 15.66 123897400
2014-07-17 00:00:00+00:00 15.2 15.48 15.13 15.45 114839900
2014-07-18 00:00:00+00:00 15.49 15.5 15.25 15.27 74859200
2014-07-21 00:00:00+00:00 15.52 15.55 15.36 15.42 61802200
2014-07-22 00:00:00+00:00 15.52 15.62 15.47 15.59 58331100
2014-07-23 00:00:00+00:00 15.52 15.63 15.51 15.52 47253300
2014-07-24 00:00:00+00:00 15.62 15.64 15.55 15.56 46549800
2014-07-25 00:00:00+00:00 15.59 15.63 15.55 15.585 35627200
2014-07-28 00:00:00+00:00 15.5 15.61 15.46 15.585 39313000
2014-07-29 00:00:00+00:00 15.34 15.53 15.34 15.52 51480500
2014-07-30 00:00:00+00:00 15.58 15.665 15.31 15.43 83243100
2014-07-31 00:00:00+00:00 15.25 15.55 15.25 15.44 70677300
2014-08-01 00:00:00+00:00 14.98 15.39 14.84 15.18 115978900
2014-08-04 00:00:00+00:00 15.05 15.12 14.98 15.065 51954700
2014-08-05 00:00:00+00:00 15 15.2 14.9 15.01 65071500
2014-08-06 00:00:00+00:00 15.2 15.36 15.14 15.14 95977900
2014-08-07 00:00:00+00:00 15.12 15.44 15.09 15.44 80058300
2014-08-08 00:00:00+00:00 15.2 15.2 14.99 15.08 54495000
2014-08-11 00:00:00+00:00 15.22 15.27 15.15 15.26 41545100
2014-08-12 00:00:00+00:00 15.21 15.3 15.15 15.18 33674400
2014-08-13 00:00:00+00:00 15.25 15.29 15.2 15.25 34394600
2014-08-14 00:00:00+00:00 15.32 15.32 15.26 15.26 29934600
2014-08-15 00:00:00+00:00 15.22 15.41 15.14 15.34 61535500
2014-08-18 00:00:00+00:00 15.45 15.45 15.27 15.28 54968600
2014-08-19 00:00:00+00:00 15.45 15.65 15.44 15.52 44825500
2014-08-20 00:00:00+00:00 15.52 15.633 15.37 15.41 57825000
2014-08-21 00:00:00+00:00 16.16 16.219 15.62 15.69 177294400
2014-08-22 00:00:00+00:00 16.129 16.29 16.05 16.16 107641800
2014-08-25 00:00:00+00:00 16.29 16.4 16.2 16.27 89396500
2014-08-26 00:00:00+00:00 16.329 16.46 16.32 16.34 73323400
2014-08-27 00:00:00+00:00 16.2 16.39 16.14 16.37 63061800
2014-08-28 00:00:00+00:00 16.01 16.1 15.99 16.094 62170400
2014-08-29 00:00:00+00:00 16.09 16.14 16.02 16.045 50106600
2014-09-02 00:00:00+00:00 16.27 16.28 16.059 16.14 59400400
2014-09-03 00:00:00+00:00 16.1 16.28 16.04 16.27 67418900
2014-09-04 00:00:00+00:00 16.11 16.27 16.04 16.094 56378600
2014-09-05 00:00:00+00:00 16.02 16.07 15.9 16.05 80974900
2014-09-08 00:00:00+00:00 16.35 16.37 16.149 16.19 99411200
2014-09-09 00:00:00+00:00 16.14 16.26 16.1 16.25 82428900
2014-09-10 00:00:00+00:00 16.36 16.4 16.129 16.155 75656100
2014-09-11 00:00:00+00:00 16.57 16.63 16.3 16.32 106598100
2014-09-12 00:00:00+00:00 16.79 16.83 16.61 16.62 117118300
2014-09-15 00:00:00+00:00 16.74 16.93 16.62 16.795 87306900
2014-09-16 00:00:00+00:00 16.71 16.84 16.67 16.67 64845600
2014-09-17 00:00:00+00:00 16.77 16.93 16.68 16.71 82813200
2014-09-18 00:00:00+00:00 17.04 17.15 16.87 16.88 110396000
2014-09-19 00:00:00+00:00 16.95 17.17 16.88 17.15 88215900
2014-09-22 00:00:00+00:00 17.03 17.17 16.99 17.09 108394117
2014-09-23 00:00:00+00:00 17.05 17.2 17.03 17.05 91178038
2014-09-24 00:00:00+00:00 17.18 17.19 16.97 17.12 85925116
2014-09-25 00:00:00+00:00 16.85 17.18 16.85 17.155 102101191
2014-09-26 00:00:00+00:00 17.03 17.05 16.89 16.91 66785876
2014-09-29 00:00:00+00:00 17.01 17.05 16.88 16.91 68015923
2014-09-30 00:00:00+00:00 17.05 17.11 16.97 17.08 82143402
2014-10-01 00:00:00+00:00 16.82 17.09 16.8 17.07 91190551
2014-10-02 00:00:00+00:00 16.88 16.99 16.63 16.86 118018774
2014-10-03 00:00:00+00:00 17.29 17.3 17.06 17.11 110782588
2014-10-06 00:00:00+00:00 17.29 17.41 17.22 17.37 66212066
2014-10-07 00:00:00+00:00 16.88 17.2 16.88 17.18 91396327
2014-10-08 00:00:00+00:00 17.12 17.12 16.72 16.88 101178019
2014-10-09 00:00:00+00:00 16.59 17.11 16.55 17.04 121286578
2014-10-10 00:00:00+00:00 16.48 16.77 16.36 16.52 129458091
2014-10-13 00:00:00+00:00 16.4 16.67 16.4 16.48 92674130
2014-10-14 00:00:00+00:00 16.52 16.63 16.36 16.51 97451881
2014-10-15 00:00:00+00:00 15.76 16.239 15.43 16.23 216591804
2014-10-16 00:00:00+00:00 16.079 16.25 15.52 15.61 148136470
2014-10-17 00:00:00+00:00 16.21 16.41 16.16 16.25 94339180
2014-10-20 00:00:00+00:00 16.26 16.329 16.16 16.2 76446700
2014-10-21 00:00:00+00:00 16.6 16.61 16.309 16.43 78226090
2014-10-22 00:00:00+00:00 16.4 16.7 16.37 16.59 85160035
2014-10-23 00:00:00+00:00 16.6 16.73 16.515 16.58 68367854
2014-10-24 00:00:00+00:00 16.72 16.72 16.56 16.63 41904957
2014-10-27 00:00:00+00:00 16.59 16.69 16.5 16.68 51484734
2014-10-28 00:00:00+00:00 16.8 16.8 16.61 16.62 71850839
2014-10-29 00:00:00+00:00 16.99 17.02 16.71 16.77 99875657
2014-10-30 00:00:00+00:00 17.03 17.12 16.84 16.96 72542662
2014-10-31 00:00:00+00:00 17.16 17.22 17.1 17.17 82788923
2014-11-03 00:00:00+00:00 17.27 17.35 17.1 17.18 64048210
2014-11-04 00:00:00+00:00 17.21 17.3 17.02 17.22 53463983
2014-11-05 00:00:00+00:00 17.34 17.37 17.21 17.32 58562613
2014-11-06 00:00:00+00:00 17.36 17.4 17.28 17.34 57175259
2014-11-07 00:00:00+00:00 17.36 17.38 17.22 17.31 53907826
2014-11-10 00:00:00+00:00 17.37 17.4 17.3 17.36 54224010
2014-11-11 00:00:00+00:00 17.32 17.46 17.3 17.37 62236029
2014-11-12 00:00:00+00:00 17.29 17.3 17.07 17.24 48641530
2014-11-13 00:00:00+00:00 17.22 17.3 17.12 17.285 50586273
2014-11-14 00:00:00+00:00 17.14 17.25 17.1 17.165 41116439
2014-11-17 00:00:00+00:00 17.09 17.14 16.97 17.06 43359749
2014-11-18 00:00:00+00:00 17.14 17.22 17.06 17.07 39490695
2014-11-19 00:00:00+00:00 17.06 17.15 17.01 17.145 49409552
2014-11-20 00:00:00+00:00 17 17.01 16.83 16.96 49641573
2014-11-21 00:00:00+00:00 17.12 17.19 17.06 17.15 62980975
2014-11-24 00:00:00+00:00 17.18 17.28 17.08 17.16 52762793
2014-11-25 00:00:00+00:00 17.1 17.26 17.08 17.23 44786324
2014-11-26 00:00:00+00:00 17.11 17.15 17.04 17.12 27358440
2014-11-28 00:00:00+00:00 17.04 17.15 17.03 17.07 27834925
2014-12-01 00:00:00+00:00 16.79 16.94 16.73 16.92 61173148
2014-12-02 00:00:00+00:00 17.15 17.15 16.86 16.87 62945173
2014-12-03 00:00:00+00:00 17.29 17.3 17.06 17.1 70104217
2014-12-04 00:00:00+00:00 17.21 17.34 17.13 17.24 49761358
2014-12-05 00:00:00+00:00 17.68 17.71 17.35 17.41 131823990
2014-12-08 00:00:00+00:00 17.66 17.87 17.51 17.66 100987510
2014-12-09 00:00:00+00:00 17.56 17.59 17.17 17.175 95939373
2014-12-10 00:00:00+00:00 17.38 17.68 17.37 17.48 103574081
2014-12-11 00:00:00+00:00 17.47 17.65 17.44 17.44 80139049
2014-12-12 00:00:00+00:00 17.13 17.45 17.13 17.37 88836682
2014-12-15 00:00:00+00:00 16.85 17.31 16.76 17.3 101305256
2014-12-16 00:00:00+00:00 16.72 17.19 16.59 16.74 96000647
2014-12-17 00:00:00+00:00 17.26 17.27 16.82 16.87 92164082
2014-12-18 00:00:00+00:00 17.53 17.53 17.34 17.47 79264648
2014-12-19 00:00:00+00:00 17.62 17.7 17.49 17.49 105588213
2014-12-22 00:00:00+00:00 17.71 17.73 17.55 17.65 70912446
2014-12-23 00:00:00+00:00 17.93 17.99 17.78 17.83 94340090
2014-12-24 00:00:00+00:00 17.98 18.1 17.92 17.965 35091254
2014-12-26 00:00:00+00:00 17.98 18.05 17.95 18.02 34324407
2014-12-29 00:00:00+00:00 18.11 18.19 17.91 17.97 58370887
2014-12-30 00:00:00+00:00 18.13 18.18 18.01 18.04 41433523
2014-12-31 00:00:00+00:00 17.89 18.21 17.89 18.19 57819572
2015-01-02 00:00:00+00:00 17.9 18.03 17.68 17.99 48915838
2015-01-05 00:00:00+00:00 17.38 17.81 17.29 17.785 105605485
2015-01-06 00:00:00+00:00 16.86 17.44 16.78 17.42 144847119
2015-01-07 00:00:00+00:00 16.94 17.18 16.87 17.14 104563015
2015-01-08 00:00:00+00:00 17.29 17.34 17.1 17.16 73310178
2015-01-09 00:00:00+00:00 16.98 17.38 16.95 17.38 82933457
2015-01-12 00:00:00+00:00 16.68 17.03 16.66 17.025 92704852
2015-01-13 00:00:00+00:00 16.45 16.89 16.32 16.82 102036015
2015-01-14 00:00:00+00:00 16.04 16.215 15.77 16 164685486
2015-01-15 00:00:00+00:00 15.2 15.76 15.15 15.59 193140606
2015-01-16 00:00:00+00:00 15.38 15.4 14.97 15.16 149802031
2015-01-20 00:00:00+00:00 15.26 15.63 15.2 15.59 123048254
2015-01-21 00:00:00+00:00 15.41 15.57 15.15 15.28 100658196
2015-01-22 00:00:00+00:00 16.09 16.19 15.43 15.55 183015219
2015-01-23 00:00:00+00:00 15.73 16.19 15.73 16.04 103268518
2015-01-26 00:00:00+00:00 15.85 15.93 15.7 15.72 69948779
2015-01-27 00:00:00+00:00 15.63 15.79 15.48 15.55 86184227
2015-01-28 00:00:00+00:00 15.2 15.73 15.18 15.72 105298688
2015-01-29 00:00:00+00:00 15.43 15.49 15.2 15.31 76055669
2015-01-30 00:00:00+00:00 15.15 15.47 15.15 15.23 99755260
2015-02-02 00:00:00+00:00 15.46 15.49 15.12 15.27 101557039
2015-02-03 00:00:00+00:00 15.89 15.93 15.61 15.62 105159425
2015-02-04 00:00:00+00:00 15.79 16 15.75 15.79 83814535
2015-02-05 00:00:00+00:00 15.97 16.09 15.9 15.98 92166071
2015-02-06 00:00:00+00:00 16.49 16.75 16.219 16.3 160764889
2015-02-09 00:00:00+00:00 16.35 16.5 16.25 16.35 95520604
2015-02-10 00:00:00+00:00 16.42 16.62 16.35 16.56 100322839
2015-02-11 00:00:00+00:00 16.36 16.45 16.21 16.329 99884358
2015-02-12 00:00:00+00:00 16.67 16.73 16.37 16.41 116394126
2015-02-13 00:00:00+00:00 16.61 16.79 16.54 16.73 93632718
2015-02-17 00:00:00+00:00 16.63 16.68 16.37 16.525 93339898
2015-02-18 00:00:00+00:00 16.3 16.56 16.239 16.55 84024165
2015-02-19 00:00:00+00:00 16.21 16.379 16.079 16.21 83591315
2015-02-20 00:00:00+00:00 16.379 16.43 16.01 16.14 89731324
2015-02-23 00:00:00+00:00 16.2 16.32 16.1 16.32 103841653
2015-02-24 00:00:00+00:00 16.379 16.55 16.25 16.27 74109596
2015-02-25 00:00:00+00:00 16.49 16.5 16.32 16.37 57200974
2015-02-26 00:00:00+00:00 16.04 16.44 15.9 16.42 161030274
2015-02-27 00:00:00+00:00 15.81 15.88 15.62 15.78 130419006
2015-03-02 00:00:00+00:00 16.01 16.03 15.715 15.79 71379858
2015-03-03 00:00:00+00:00 16.04 16.149 15.96 16.03 65627992
2015-03-04 00:00:00+00:00 15.84 15.98 15.76 15.96 78084858
2015-03-05 00:00:00+00:00 16 16.03 15.745 15.92 69509958
2015-03-06 00:00:00+00:00 16.219 16.62 16.16 16.41 163207226
2015-03-09 00:00:00+00:00 16.17 16.329 16.12 16.309 72912722
2015-03-10 00:00:00+00:00 15.79 16.129 15.79 16.04 85633552
2015-03-11 00:00:00+00:00 16.11 16.16 15.87 15.92 85531260
2015-03-12 00:00:00+00:00 16.09 16.21 15.9 16.05 126376330
2015-03-13 00:00:00+00:00 16.09 16.149 15.94 16.079 88561897
2015-03-16 00:00:00+00:00 16.129 16.219 15.85 16.11 63374831
2015-03-17 00:00:00+00:00 16.09 16.12 15.96 16.07 67297458
2015-03-18 00:00:00+00:00 15.98 16.1 15.91 16.01 86363525
2015-03-19 00:00:00+00:00 15.61 15.97 15.61 15.96 109092453
2015-03-20 00:00:00+00:00 15.84 15.93 15.64 15.695 99867679
2015-03-23 00:00:00+00:00 15.72 15.88 15.72 15.82 72662013
2015-03-24 00:00:00+00:00 15.61 15.8 15.61 15.73 76599529
2015-03-25 00:00:00+00:00 15.41 15.65 15.4 15.63 88466042
2015-03-26 00:00:00+00:00 15.42 15.54 15.26 15.385 76699386
2015-03-27 00:00:00+00:00 15.31 15.49 15.27 15.48 72913154
2015-03-30 00:00:00+00:00 15.52 15.61 15.4 15.42 70819151
2015-03-31 00:00:00+00:00 15.39 15.52 15.38 15.5 61333548
2015-04-01 00:00:00+00:00 15.41 15.46 15.25 15.42 73153854
2015-04-02 00:00:00+00:00 15.54 15.62 15.43 15.43 50443508
2015-04-06 00:00:00+00:00 15.51 15.6 15.34 15.39 51161893
2015-04-07 00:00:00+00:00 15.46 15.65 15.45 15.53 49566392
2015-04-08 00:00:00+00:00 15.61 15.74 15.5 15.51 71732476
2015-04-09 00:00:00+00:00 15.71 15.76 15.53 15.63 44719259
2015-04-10 00:00:00+00:00 15.72 15.79 15.6 15.69 43801947
2015-04-13 00:00:00+00:00 15.8 15.86 15.74 15.78 49191244
2015-04-14 00:00:00+00:00 15.82 15.95 15.71 15.88 84385405
2015-04-15 00:00:00+00:00 15.64 15.85 15.59 15.7 124479058
2015-04-16 00:00:00+00:00 15.79 15.94 15.58 15.64 105591004
2015-04-17 00:00:00+00:00 15.56 15.75 15.5 15.71 88852590
2015-04-20 00:00:00+00:00 15.57 15.69 15.56 15.62 53649399
2015-04-21 00:00:00+00:00 15.5 15.64 15.43 15.6 64033133
2015-04-22 00:00:00+00:00 15.74 15.83 15.49 15.55 73075265
2015-04-23 00:00:00+00:00 15.69 15.8 15.68 15.72 50234512
2015-04-24 00:00:00+00:00 15.64 15.75 15.61 15.71 40766109
2015-04-27 00:00:00+00:00 15.56 15.76 15.56 15.63 73436466
2015-04-28 00:00:00+00:00 15.65 15.71 15.5 15.58 58076725
2015-04-29 00:00:00+00:00 15.98 16.04 15.57 15.6 134438101
2015-04-30 00:00:00+00:00 15.93 16.05 15.795 16 78201256
2015-05-01 00:00:00+00:00 16.11 16.149 15.92 16 78171362
2015-05-04 00:00:00+00:00 16.44 16.45 16.12 16.14 76504381
2015-05-05 00:00:00+00:00 16.35 16.61 16.329 16.42 106053204
2015-05-06 00:00:00+00:00 16.29 16.49 16.079 16.36 96396889
2015-05-07 00:00:00+00:00 16.239 16.329 16.09 16.23 73740256
2015-05-08 00:00:00+00:00 16.45 16.46 16.219 16.32 86321384
2015-05-11 00:00:00+00:00 16.49 16.59 16.43 16.45 56185501
2015-05-12 00:00:00+00:00 16.43 16.48 16.35 16.46 59589310
2015-05-13 00:00:00+00:00 16.47 16.52 16.36 16.44 47330638
2015-05-14 00:00:00+00:00 16.52 16.59 16.45 16.53 55380166
2015-05-15 00:00:00+00:00 16.35 16.52 16.3 16.52 54868077
2015-05-18 00:00:00+00:00 16.51 16.54 16.309 16.309 51047229
2015-05-19 00:00:00+00:00 16.77 16.78 16.58 16.59 89115475
2015-05-20 00:00:00+00:00 16.74 16.85 16.63 16.78 67640605
2015-05-21 00:00:00+00:00 16.73 16.75 16.56 16.71 52045653
2015-05-22 00:00:00+00:00 16.75 16.8 16.705 16.72 47032906
2015-05-26 00:00:00+00:00 16.5 16.73 16.43 16.73 99306419
2015-05-27 00:00:00+00:00 16.74 16.75 16.54 16.56 69999158
2015-05-28 00:00:00+00:00 16.67 16.73 16.58 16.72 60969979
2015-05-29 00:00:00+00:00 16.5 16.67 16.47 16.66 74246895
2015-06-01 00:00:00+00:00 16.55 16.64 16.47 16.58 62912633
2015-06-02 00:00:00+00:00 16.72 16.76 16.5 16.52 65498460
2015-06-03 00:00:00+00:00 16.93 17.02 16.74 16.79 89614868
2015-06-04 00:00:00+00:00 16.78 16.98 16.72 16.87 61025453
2015-06-05 00:00:00+00:00 17.19 17.35 16.965 17.04 119087408
2015-06-08 00:00:00+00:00 17.08 17.33 17.05 17.27 69749600
2015-06-09 00:00:00+00:00 17.31 17.35 17 17.05 82104858
2015-06-10 00:00:00+00:00 17.59 17.62 17.36 17.39 102643147
2015-06-11 00:00:00+00:00 17.49 17.675 17.44 17.62 77191527
2015-06-12 00:00:00+00:00 17.49 17.58 17.36 17.45 53149442
2015-06-15 00:00:00+00:00 17.47 17.5 17.25 17.33 68053373
2015-06-16 00:00:00+00:00 17.55 17.56 17.37 17.46 47291181
2015-06-17 00:00:00+00:00 17.37 17.595 17.3 17.59 88712103
2015-06-18 00:00:00+00:00 17.38 17.4 17.22 17.35 97574677
2015-06-19 00:00:00+00:00 17.17 17.39 17.12 17.28 81165957
2015-06-22 00:00:00+00:00 17.47 17.53 17.35 17.36 58447195
2015-06-23 00:00:00+00:00 17.67 17.72 17.56 17.56 63865708
2015-06-24 00:00:00+00:00 17.49 17.69 17.45 17.53 59052661
2015-06-25 00:00:00+00:00 17.37 17.62 17.35 17.57 59704817
2015-06-26 00:00:00+00:00 17.41 17.52 17.35 17.48 62897209
2015-06-29 00:00:00+00:00 16.89 17.25 16.86 17.13 98515800
2015-06-30 00:00:00+00:00 17.02 17.13 16.85 17.08 89039753
2015-07-01 00:00:00+00:00 17.22 17.31 17.09 17.25 62317352
2015-07-02 00:00:00+00:00 17.03 17.21 16.89 17.16 64736475
2015-07-06 00:00:00+00:00 16.94 17.01 16.71 16.78 58666946
2015-07-07 00:00:00+00:00 16.69 16.93 16.34 16.9 115399859
2015-07-08 00:00:00+00:00 16.25 16.5 16.219 16.42 83032960
2015-07-09 00:00:00+00:00 16.48 16.72 16.45 16.53 81534426
2015-07-10 00:00:00+00:00 16.7 16.84 16.66 16.795 72941057
2015-07-13 00:00:00+00:00 17.02 17.05 16.9 16.97 67934294
2015-07-14 00:00:00+00:00 17.13 17.15 16.86 16.93 69933168
2015-07-15 00:00:00+00:00 17.68 17.85 17.49 17.53 173099934
2015-07-16 00:00:00+00:00 17.95 18.07 17.86 17.91 117054318
2015-07-17 00:00:00+00:00 18.1 18.16 17.89 17.96 96860681
2015-07-20 00:00:00+00:00 18.12 18.23 18.08 18.15 71909590
2015-07-21 00:00:00+00:00 18.08 18.17 18.02 18.11 63684538
2015-07-22 00:00:00+00:00 18.45 18.48 18.03 18.03 110097735
2015-07-23 00:00:00+00:00 18.18 18.43 18.08 18.39 103650135
2015-07-24 00:00:00+00:00 17.9 18.23 17.84 18.2 86296535
2015-07-27 00:00:00+00:00 17.67 17.78 17.5 17.66 93231917
2015-07-28 00:00:00+00:00 17.88 17.89 17.6 17.79 82154817
2015-07-29 00:00:00+00:00 18.16 18.2 17.92 17.95 96388569
2015-07-30 00:00:00+00:00 18.13 18.29 18.05 18.2 65660338
2015-07-31 00:00:00+00:00 17.88 18.06 17.86 18.03 67176946
2015-08-03 00:00:00+00:00 17.77 17.965 17.64 17.91 62341245
2015-08-04 00:00:00+00:00 17.8 17.93 17.71 17.79 69741507
2015-08-05 00:00:00+00:00 17.87 18.045 17.8 17.93 65044706
2015-08-06 00:00:00+00:00 17.81 17.975 17.77 17.91 47582061
2015-08-07 00:00:00+00:00 17.75 18.07 17.6 17.92 71020703
2015-08-10 00:00:00+00:00 18.04 18.05 17.86 17.87 65765231
2015-08-11 00:00:00+00:00 17.79 17.97 17.72 17.85 65868585
2015-08-12 00:00:00+00:00 17.52 17.645 17.02 17.62 143433552
2015-08-13 00:00:00+00:00 17.62 17.69 17.44 17.56 67406457
2015-08-14 00:00:00+00:00 17.7 17.7 17.54 17.57 52710258
2015-08-17 00:00:00+00:00 17.77 17.81 17.56 17.61 42309963
2015-08-18 00:00:00+00:00 17.69 17.9 17.65 17.72 50346194
2015-08-19 00:00:00+00:00 17.46 17.69 17.44 17.55 78522838
2015-08-20 00:00:00+00:00 16.72 17.2 16.72 17.19 149185431
2015-08-21 00:00:00+00:00 16.1 16.66 16.1 16.195 148684325
2015-08-24 00:00:00+00:00 15.29 15.98 14.6 15.02 214559086
2015-08-25 00:00:00+00:00 15.26 16.204 15.25 16.2 194227130
2015-08-26 00:00:00+00:00 16.059 16.059 15.335 15.81 173838973
2015-08-27 00:00:00+00:00 16.44 16.45 16.09 16.36 124129428
2015-08-28 00:00:00+00:00 16.36 16.44 16.2 16.4 75850400
2015-08-31 00:00:00+00:00 16.34 16.375 16.18 16.329 66685093
2015-09-01 00:00:00+00:00 15.58 16.05 15.46 15.95 118974368
2015-09-02 00:00:00+00:00 15.85 15.86 15.57 15.81 74737988
2015-09-03 00:00:00+00:00 15.94 16.19 15.87 15.97 77175757
2015-09-04 00:00:00+00:00 15.65 15.86 15.5 15.79 104539500
2015-09-08 00:00:00+00:00 16.16 16.16 15.9 15.96 73939623
2015-09-09 00:00:00+00:00 15.9 16.46 15.83 16.37 70466139
2015-09-10 00:00:00+00:00 16.04 16.11 15.81 15.87 71085643
2015-09-11 00:00:00+00:00 16.04 16.059 15.9 15.99 54275460
2015-09-14 00:00:00+00:00 15.96 16.07 15.9 15.97 50640946
2015-09-15 00:00:00+00:00 16.309 16.39 16.01 16.04 62327105
2015-09-16 00:00:00+00:00 16.329 16.39 16.09 16.35 83721453
2015-09-17 00:00:00+00:00 15.86 16.48 15.78 16.29 119088186
2015-09-18 00:00:00+00:00 15.56 15.71 15.5 15.68 136531198
2015-09-21 00:00:00+00:00 15.7 15.75 15.57 15.67 79081690
2015-09-22 00:00:00+00:00 15.57 15.58 15.45 15.49 92227216
2015-09-23 00:00:00+00:00 15.72 15.85 15.58 15.6 78117507
2015-09-24 00:00:00+00:00 15.55 15.62 15.4 15.535 99432185
2015-09-25 00:00:00+00:00 15.89 16.02 15.81 15.88 88567972
2015-09-28 00:00:00+00:00 15.47 15.82 15.38 15.79 86826540
2015-09-29 00:00:00+00:00 15.35 15.53 15.25 15.5 79804455
2015-09-30 00:00:00+00:00 15.58 15.6 15.32 15.55 71074169
2015-10-01 00:00:00+00:00 15.55 15.64 15.36 15.52 71471505
2015-10-02 00:00:00+00:00 15.38 15.38 14.63 15.08 181788433
2015-10-05 00:00:00+00:00 15.69 15.79 15.44 15.45 75293923
2015-10-06 00:00:00+00:00 15.69 15.82 15.57 15.68 67445122
2015-10-07 00:00:00+00:00 15.75 15.93 15.57 15.78 66997033
2015-10-08 00:00:00+00:00 15.75 15.78 15.51 15.72 79056883
2015-10-09 00:00:00+00:00 15.58 15.82 15.5 15.75 76667097
2015-10-12 00:00:00+00:00 15.52 15.62 15.43 15.6 50725777
2015-10-13 00:00:00+00:00 15.52 15.59 15.4 15.45 73689847
2015-10-14 00:00:00+00:00 15.64 15.92 15.55 15.77 122207202
2015-10-15 00:00:00+00:00 16.19 16.28 15.72 15.81 127028772
2015-10-16 00:00:00+00:00 16.12 16.29 16.03 16.285 66966667
2015-10-19 00:00:00+00:00 16.14 16.21 15.98 16 62953541
2015-10-20 00:00:00+00:00 16.2 16.29 16.1 16.16 50797790
2015-10-21 00:00:00+00:00 15.9 16.29 15.9 16.26 64825296
2015-10-22 00:00:00+00:00 16.16 16.2 15.995 16 74099343
2015-10-23 00:00:00+00:00 16.52 16.55 16.25 16.29 82285833
2015-10-26 00:00:00+00:00 16.51 16.55 16.34 16.52 67702832
2015-10-27 00:00:00+00:00 16.4 16.47 16.3 16.4 56147451
2015-10-28 00:00:00+00:00 17.28 17.31 16.445 16.45 147672009
2015-10-29 00:00:00+00:00 17.09 17.44 16.97 17.16 95510244
2015-10-30 00:00:00+00:00 16.78 17.18 16.76 17.18 87645788
2015-11-02 00:00:00+00:00 17.06 17.14 16.87 16.9 56894296
2015-11-03 00:00:00+00:00 17.18 17.26 16.99 17.01 66096236
2015-11-04 00:00:00+00:00 17.01 17.31 16.96 17.3 78688573
2015-11-05 00:00:00+00:00 17.31 17.37 17.03 17.03 85477288
2015-11-06 00:00:00+00:00 17.95 18.09 17.76 17.84 157987106
2015-11-09 00:00:00+00:00 17.68 18.08 17.56 18.03 129842355
2015-11-10 00:00:00+00:00 17.85 17.94 17.56 17.63 59845097
2015-11-11 00:00:00+00:00 17.75 17.99 17.68 17.99 59235017
2015-11-12 00:00:00+00:00 17.37 17.63 17.35 17.58 77712229
2015-11-13 00:00:00+00:00 17.2 17.31 17.09 17.13 103497332
2015-11-16 00:00:00+00:00 17.43 17.46 17.01 17.13 70377293
2015-11-17 00:00:00+00:00 17.42 17.6 17.34 17.5 70480628
2015-11-18 00:00:00+00:00 17.84 17.87 17.43 17.43 85633743
2015-11-19 00:00:00+00:00 17.69 17.85 17.62 17.78 51812145
2015-11-20 00:00:00+00:00 17.65 17.833 17.59 17.81 56264227
2015-11-23 00:00:00+00:00 17.47 17.73 17.46 17.62 50350218
2015-11-24 00:00:00+00:00 17.47 17.57 17.25 17.26 58724161
2015-11-25 00:00:00+00:00 17.44 17.57 17.41 17.51 34926498
2015-11-27 00:00:00+00:00 17.48 17.5 17.33 17.46 22937528
2015-11-30 00:00:00+00:00 17.43 17.58 17.42 17.48 61852667
2015-12-01 00:00:00+00:00 17.81 17.81 17.48 17.52 74298721
2015-12-02 00:00:00+00:00 17.62 17.89 17.55 17.88 76857387
2015-12-03 00:00:00+00:00 17.3 17.765 17.25 17.68 95781361
2015-12-04 00:00:00+00:00 17.8 17.83 17.38 17.44 102999510
2015-12-07 00:00:00+00:00 17.54 17.8 17.44 17.79 81623517
2015-12-08 00:00:00+00:00 17.19 17.459 17.13 17.39 84772872
2015-12-09 00:00:00+00:00 17.1 17.38 16.87 17.11 82418050
2015-12-10 00:00:00+00:00 17.2 17.41 16.96 17.15 68375559
2015-12-11 00:00:00+00:00 16.73 17.06 16.64 16.97 91451006
2015-12-14 00:00:00+00:00 16.8 16.89 16.495 16.76 121067256
2015-12-15 00:00:00+00:00 17.42 17.49 16.99 17.02 99737154
2015-12-16 00:00:00+00:00 17.75 17.78 17.23 17.65 171512739
2015-12-17 00:00:00+00:00 17.3 17.83 17.3 17.8 97265726
2015-12-18 00:00:00+00:00 16.76 17.265 16.76 17.19 136520314
2015-12-21 00:00:00+00:00 16.97 17.03 16.77 16.98 65133240
2015-12-22 00:00:00+00:00 17.08 17.11 16.85 17.05 56144957
2015-12-23 00:00:00+00:00 17.34 17.34 17.095 17.16 65700709
2015-12-24 00:00:00+00:00 17.27 17.38 17.22 17.32 29373415
2015-12-28 00:00:00+00:00 17.13 17.23 16.98 17.22 41759993
2015-12-29 00:00:00+00:00 17.28 17.35 17.16 17.25 45628449
2015-12-30 00:00:00+00:00 17.05 17.24 17.04 17.2 35035518
2015-12-31 00:00:00+00:00 16.83 17.07 16.83 17.01 47106760
2016-01-04 00:00:00+00:00 16.43 16.49 16.25 16.45 114855342
2016-01-05 00:00:00+00:00 16.43 16.59 16.23 16.52 66591885
2016-01-06 00:00:00+00:00 16.079 16.29 16.02 16.19 102669915
2016-01-07 00:00:00+00:00 15.5 15.9 15.44 15.73 116188238
2016-01-08 00:00:00+00:00 15.2 15.94 15.16 15.94 124670826
2016-01-11 00:00:00+00:00 15.31 15.37 14.94 15.26 104581569
2016-01-12 00:00:00+00:00 15.31 15.58 15.06 15.54 99949556
2016-01-13 00:00:00+00:00 14.9 15.52 14.85 15.47 119412972
2016-01-14 00:00:00+00:00 14.99 15.2 14.65 15.01 125809900
2016-01-15 00:00:00+00:00 14.46 14.66 14.13 14.41 172228199
2016-01-19 00:00:00+00:00 14.24 14.79 14.01 14.69 185641283
2016-01-20 00:00:00+00:00 13.69 14 13.27 13.79 249990820
2016-01-21 00:00:00+00:00 13.36 13.84 13.25 13.67 188802260
2016-01-22 00:00:00+00:00 13.56 13.72 13.47 13.65 169961438
2016-01-25 00:00:00+00:00 12.96 13.55 12.94 13.54 186248993
2016-01-26 00:00:00+00:00 13.31 13.35 13.04 13.07 124183841
2016-01-27 00:00:00+00:00 13.36 13.73 13.19 13.2 123984025
2016-01-28 00:00:00+00:00 13.53 13.705 13.26 13.59 110951618
2016-01-29 00:00:00+00:00 14.14 14.15 13.59 13.66 159925361
2016-02-01 00:00:00+00:00 13.96 14.09 13.8 14.05 105699706
2016-02-02 00:00:00+00:00 13.23 13.745 13.13 13.74 146990214
2016-02-03 00:00:00+00:00 13.03 13.29 12.52 13.28 257591135
2016-02-04 00:00:00+00:00 13.25 13.59 12.885 12.89 179319106
2016-02-05 00:00:00+00:00 12.95 13.39 12.89 13.32 125785587
2016-02-08 00:00:00+00:00 12.27 12.7 12.13 12.67 221271714
2016-02-09 00:00:00+00:00 12.2 12.43 11.96 11.99 247227500
2016-02-10 00:00:00+00:00 11.98 12.54 11.91 12.42 227371472
2016-02-11 00:00:00+00:00 11.16 11.55 10.99 11.46 374686686
2016-02-12 00:00:00+00:00 11.95 12.03 11.4 11.48 245320697
2016-02-16 00:00:00+00:00 12.25 12.39 12.11 12.38 194138587
2016-02-17 00:00:00+00:00 12.56 12.69 12.47 12.57 221671695
2016-02-18 00:00:00+00:00 12.24 12.74 12.1 12.71 161492893
2016-02-19 00:00:00+00:00 12.13 12.24 11.98 12.22 120974304
2016-02-22 00:00:00+00:00 12.54 12.54 12.36 12.38 88785977
2016-02-23 00:00:00+00:00 12.16 12.475 12.1 12.47 104401671
2016-02-24 00:00:00+00:00 12.13 12.15 11.65 11.96 159423970
2016-02-25 00:00:00+00:00 12.32 12.33 12.09 12.14 103003028
2016-02-26 00:00:00+00:00 12.7 12.95 12.4 12.49 160262798
2016-02-29 00:00:00+00:00 12.52 12.86 12.51 12.7 123480562
2016-03-01 00:00:00+00:00 13.19 13.21 12.62 12.64 153473420
2016-03-02 00:00:00+00:00 13.41 13.51 13.13 13.2 139562706
2016-03-03 00:00:00+00:00 13.5 13.55 13.25 13.38 105031428
2016-03-04 00:00:00+00:00 13.54 13.89 13.49 13.76 171338150
2016-03-07 00:00:00+00:00 13.53 13.62 13.37 13.45 97565818
2016-03-08 00:00:00+00:00 13.06 13.465 13.05 13.4 124883295
2016-03-09 00:00:00+00:00 13.14 13.27 13 13.17 92198484
2016-03-10 00:00:00+00:00 13.27 13.4 13 13.23 123467835
2016-03-11 00:00:00+00:00 13.79 13.84 13.37 13.44 120366396
2016-03-14 00:00:00+00:00 13.64 13.74 13.49 13.72 87695689
2016-03-15 00:00:00+00:00 13.57 13.6 13.39 13.51 80403985
2016-03-16 00:00:00+00:00 13.31 13.81 13.09 13.51 149645421
2016-03-17 00:00:00+00:00 13.4 13.475 13.05 13.22 122129018
2016-03-18 00:00:00+00:00 13.79 13.88 13.55 13.68 146530926
2016-03-21 00:00:00+00:00 13.84 14.03 13.72 13.8 104633493
2016-03-22 00:00:00+00:00 13.76 13.84 13.63 13.67 83196383
2016-03-23 00:00:00+00:00 13.62 13.83 13.6 13.77 95747961
2016-03-24 00:00:00+00:00 13.68 13.69 13.27 13.41 94772909
2016-03-28 00:00:00+00:00 13.62 13.74 13.54 13.73 54455597
2016-03-29 00:00:00+00:00 13.42 13.54 13.26 13.54 101606836
2016-03-30 00:00:00+00:00 13.48 13.66 13.46 13.49 85876825
2016-03-31 00:00:00+00:00 13.52 13.71 13.45 13.49 79560389
2016-04-01 00:00:00+00:00 13.56 13.65 13.32 13.47 73092379
2016-04-04 00:00:00+00:00 13.51 13.65 13.41 13.54 59638541
2016-04-05 00:00:00+00:00 13.19 13.34 13.16 13.3 78398301
2016-04-06 00:00:00+00:00 13.27 13.3 13.11 13.2 62509221
2016-04-07 00:00:00+00:00 12.85 13.205 12.75 13.15 105619597
2016-04-08 00:00:00+00:00 12.88 13.14 12.86 13.03 76527808
2016-04-11 00:00:00+00:00 12.97 13.09 12.88 12.92 79266678
2016-04-12 00:00:00+00:00 13.27 13.33 12.93 13 100558608
2016-04-13 00:00:00+00:00 13.79 13.85 13.53 13.55 138454434
2016-04-14 00:00:00+00:00 14.14 14.28 13.7 13.71 179342548
2016-04-15 00:00:00+00:00 14 14.29 13.95 14.27 87934720
2016-04-18 00:00:00+00:00 14.17 14.24 13.82 13.85 81952829
2016-04-19 00:00:00+00:00 14.45 14.54 14.24 14.26 107054754
2016-04-20 00:00:00+00:00 14.93 14.94 14.53 14.56 115452870
2016-04-21 00:00:00+00:00 14.9 15.135 14.75 14.98 121502366
2016-04-22 00:00:00+00:00 15.11 15.14 14.865 14.865 83399543
2016-04-25 00:00:00+00:00 14.96 15.14 14.81 15.02 70489478
2016-04-26 00:00:00+00:00 15.09 15.14 14.8 15.02 84745857
2016-04-27 00:00:00+00:00 15.02 15.3 14.98 15.02 115753717
2016-04-28 00:00:00+00:00 14.79 15.08 14.76 14.92 78536135
2016-04-29 00:00:00+00:00 14.56 14.85 14.43 14.73 122108474
2016-05-02 00:00:00+00:00 14.77 14.78 14.4 14.58 68882814
2016-05-03 00:00:00+00:00 14.36 14.51 14.15 14.51 110487308
2016-05-04 00:00:00+00:00 14.13 14.27 14 14.09 99238480
2016-05-05 00:00:00+00:00 14.05 14.32 14 14.15 73052136
2016-05-06 00:00:00+00:00 14.11 14.14 13.8 13.83 76555048
2016-05-09 00:00:00+00:00 13.99 14.19 13.91 14.08 55342432
2016-05-10 00:00:00+00:00 14.3 14.33 14.05 14.08 59857712
2016-05-11 00:00:00+00:00 14.2 14.5 14.19 14.25 68041516
2016-05-12 00:00:00+00:00 14.14 14.47 14.05 14.3 72550966
2016-05-13 00:00:00+00:00 13.88 14.37 13.87 14.15 86141548
2016-05-16 00:00:00+00:00 13.93 14.02 13.8 13.82 57848601
2016-05-17 00:00:00+00:00 14.01 14.13 13.82 13.89 76237641
2016-05-18 00:00:00+00:00 14.69 14.75 14.01 14.02 151871694
2016-05-19 00:00:00+00:00 14.53 14.865 14.43 14.595 106156946
2016-05-20 00:00:00+00:00 14.52 14.7 14.43 14.64 81195479
2016-05-23 00:00:00+00:00 14.47 14.6 14.4 14.54 66278981
2016-05-24 00:00:00+00:00 14.68 14.75 14.53 14.6 96140736
2016-05-25 00:00:00+00:00 14.92 15.15 14.83 14.83 123568712
2016-05-26 00:00:00+00:00 14.7 14.99 14.69 14.98 65307024
2016-05-27 00:00:00+00:00 14.88 14.88 14.69 14.76 62383643
2016-05-31 00:00:00+00:00 14.79 15.05 14.73 15.03 85580697
2016-06-01 00:00:00+00:00 14.86 14.9 14.45 14.6 62596061
2016-06-02 00:00:00+00:00 14.94 14.98 14.77 14.95 64327307
2016-06-03 00:00:00+00:00 14.42 14.5 14.19 14.46 148554288
2016-06-06 00:00:00+00:00 14.52 14.7 14.37 14.44 80939209
2016-06-07 00:00:00+00:00 14.35 14.6 14.34 14.54 67447137
2016-06-08 00:00:00+00:00 14.43 14.46 14.3 14.35 58292864
2016-06-09 00:00:00+00:00 14.19 14.32 14.08 14.32 74782442
2016-06-10 00:00:00+00:00 13.83 14 13.75 13.98 90725677
2016-06-13 00:00:00+00:00 13.6 13.905 13.58 13.64 74831253
2016-06-14 00:00:00+00:00 13.26 13.726 13.17 13.56 85308733
2016-06-15 00:00:00+00:00 13.34 13.66 13.28 13.38 100619784
2016-06-16 00:00:00+00:00 13.31 13.33 13.02 13.23 90082453
2016-06-17 00:00:00+00:00 13.4 13.53 13.28 13.38 89022300
2016-06-20 00:00:00+00:00 13.54 13.85 13.51 13.74 89873995
2016-06-21 00:00:00+00:00 13.62 13.65 13.445 13.615 73730632
2016-06-22 00:00:00+00:00 13.61 13.78 13.59 13.6 77910297
2016-06-23 00:00:00+00:00 14.04 14.05 13.82 13.84 81994923
2016-06-24 00:00:00+00:00 13 13.44 12.97 13.05 231376163
2016-06-27 00:00:00+00:00 12.18 12.77 12.05 12.77 259697521
2016-06-28 00:00:00+00:00 12.7 12.72 12.35 12.57 150037397
2016-06-29 00:00:00+00:00 13.19 13.22 12.92 13.07 118484022
2016-06-30 00:00:00+00:00 13.27 13.38 13.07 13.37 125773865
2016-07-01 00:00:00+00:00 13.1 13.27 13.02 13.19 88806340
2016-07-05 00:00:00+00:00 12.74 12.94 12.63 12.93 98165899
2016-07-06 00:00:00+00:00 12.86 12.92 12.45 12.52 94953823
2016-07-07 00:00:00+00:00 13.01 13.11 12.84 12.865 101100345
2016-07-08 00:00:00+00:00 13.17 13.32 13.11 13.28 92698914
2016-07-11 00:00:00+00:00 13.21 13.4 13.2 13.29 67222146
2016-07-12 00:00:00+00:00 13.54 13.6 13.4 13.41 89042101
2016-07-13 00:00:00+00:00 13.44 13.58 13.32 13.5 74431683
2016-07-14 00:00:00+00:00 13.65 13.78 13.64 13.73 97784904
2016-07-15 00:00:00+00:00 13.66 13.79 13.52 13.78 78528450
2016-07-18 00:00:00+00:00 14.11 14.23 13.82 13.84 187014216
2016-07-19 00:00:00+00:00 14.26 14.37 14.03 14.06 96838139
2016-07-20 00:00:00+00:00 14.4 14.43 14.26 14.35 79212132
2016-07-21 00:00:00+00:00 14.27 14.47 14.26 14.43 70904379
2016-07-22 00:00:00+00:00 14.38 14.4 14.18 14.26 46604794
2016-07-25 00:00:00+00:00 14.37 14.39 14.28 14.31 46404316
2016-07-26 00:00:00+00:00 14.53 14.55 14.31 14.32 60153804
2016-07-27 00:00:00+00:00 14.63 14.7 14.45 14.57 100965373
2016-07-28 00:00:00+00:00 14.68 14.69 14.46 14.57 72049503
2016-07-29 00:00:00+00:00 14.49 14.7 14.48 14.56 63257654
2016-08-01 00:00:00+00:00 14.33 14.6 14.26 14.52 61279863
2016-08-02 00:00:00+00:00 14.13 14.45 14.09 14.28 83776290
2016-08-03 00:00:00+00:00 14.48 14.48 14.095 14.11 65263266
2016-08-04 00:00:00+00:00 14.48 14.54 14.36 14.48 46171776
2016-08-05 00:00:00+00:00 15.05 15.06 14.75 14.75 120947818
2016-08-08 00:00:00+00:00 15.13 15.17 14.92 15.08 71936030
2016-08-09 00:00:00+00:00 15.19 15.19 15.07 15.11 49973298
2016-08-10 00:00:00+00:00 14.81 15.18 14.78 15.16 79080688
2016-08-11 00:00:00+00:00 14.88 14.97 14.745 14.83 68945407
2016-08-12 00:00:00+00:00 14.91 14.91 14.7 14.77 63022911
2016-08-15 00:00:00+00:00 15.02 15.03 14.945 14.97 48592071
2016-08-16 00:00:00+00:00 15.17 15.19 14.93 14.97 80307508
2016-08-17 00:00:00+00:00 15.15 15.24 15.08 15.16 93372229
2016-08-18 00:00:00+00:00 15.16 15.24 15.06 15.11 62804324
2016-08-19 00:00:00+00:00 15.22 15.25 15.05 15.13 59800158
2016-08-22 00:00:00+00:00 15.18 15.25 15.12 15.2 60877097
2016-08-23 00:00:00+00:00 15.35 15.41 15.26 15.26 67798132
2016-08-24 00:00:00+00:00 15.4 15.5 15.36 15.37 65415804
2016-08-25 00:00:00+00:00 15.53 15.54 15.4 15.42 67624942
2016-08-26 00:00:00+00:00 15.79 15.9 15.58 15.61 126886261
2016-08-29 00:00:00+00:00 15.84 16.01 15.775 15.81 120111303
2016-08-30 00:00:00+00:00 16.19 16.23 15.8 15.87 121173617
2016-08-31 00:00:00+00:00 16.14 16.239 15.88 16.21 127664267
2016-09-01 00:00:00+00:00 15.98 16.149 15.75 16.149 128812599
2016-09-02 00:00:00+00:00 16 16.059 15.82 15.94 100972145
2016-09-06 00:00:00+00:00 15.78 16.059 15.7 16.05 96503738
2016-09-07 00:00:00+00:00 15.7 15.85 15.65 15.72 63284147
2016-09-08 00:00:00+00:00 15.86 15.94 15.68 15.77 69292664
2016-09-09 00:00:00+00:00 15.74 16.149 15.74 15.91 135182534
2016-09-12 00:00:00+00:00 15.9 15.94 15.48 15.65 109723978
2016-09-13 00:00:00+00:00 15.72 15.86 15.55 15.67 89317564
2016-09-14 00:00:00+00:00 15.63 15.84 15.56 15.67 75517820
2016-09-15 00:00:00+00:00 15.67 15.745 15.58 15.64 67883486
2016-09-16 00:00:00+00:00 15.49 15.68 15.48 15.66 90873348
2016-09-19 00:00:00+00:00 15.59 15.78 15.55 15.55 66168885
2016-09-20 00:00:00+00:00 15.6 15.76 15.51 15.74 68201707
2016-09-21 00:00:00+00:00 15.65 15.83 15.39 15.71 95464559
2016-09-22 00:00:00+00:00 15.6 15.71 15.5 15.69 75586903
2016-09-23 00:00:00+00:00 15.52 15.67 15.5 15.53 51578372
2016-09-26 00:00:00+00:00 15.09 15.44 15.02 15.4 94947767
2016-09-27 00:00:00+00:00 15.29 15.34 14.81 15.01 81992627
2016-09-28 00:00:00+00:00 15.38 15.39 15.15 15.36 66021211
2016-09-29 00:00:00+00:00 15.16 15.5 15.06 15.38 78815768
2016-09-30 00:00:00+00:00 15.65 15.73 15.17 15.26 119309179
2016-10-03 00:00:00+00:00 15.63 15.73 15.5 15.59 69864051
2016-10-04 00:00:00+00:00 15.8 16.05 15.67 15.7 98501476
2016-10-05 00:00:00+00:00 16.11 16.23 15.96 15.96 86056859
2016-10-06 00:00:00+00:00 16.219 16.23 16.035 16.17 75838739
2016-10-07 00:00:00+00:00 16.129 16.25 15.98 16.19 99597964
2016-10-10 00:00:00+00:00 16.3 16.4 16.184 16.215 64158491
2016-10-11 00:00:00+00:00 16.11 16.329 16.024 16.26 77348564
2016-10-12 00:00:00+00:00 16.03 16.239 16.01 16.09 67144049
2016-10-13 00:00:00+00:00 15.83 16.09 15.6 16.09 78636990
2016-10-14 00:00:00+00:00 16 16.23 15.94 16.149 108815580
2016-10-17 00:00:00+00:00 16.05 16.195 15.9 16.17 101795541
2016-10-18 00:00:00+00:00 16.26 16.27 16.11 16.19 71433078
2016-10-19 00:00:00+00:00 16.47 16.53 16.28 16.3 97968987
2016-10-20 00:00:00+00:00 16.56 16.63 16.41 16.45 88077661
2016-10-21 00:00:00+00:00 16.67 16.67 16.4 16.46 77921403
2016-10-24 00:00:00+00:00 16.77 16.8 16.61 16.75 68237544
2016-10-25 00:00:00+00:00 16.72 16.82 16.68 16.71 49342342
2016-10-26 00:00:00+00:00 16.87 16.87 16.62 16.64 58690167
2016-10-27 00:00:00+00:00 16.91 17.1 16.864 16.95 97937253
2016-10-28 00:00:00+00:00 16.68 16.98 16.5 16.95 115090796
2016-10-31 00:00:00+00:00 16.5 16.73 16.5 16.68 70352567
2016-11-01 00:00:00+00:00 16.61 16.75 16.329 16.56 88557065
2016-11-02 00:00:00+00:00 16.48 16.56 16.28 16.46 119136779
2016-11-03 00:00:00+00:00 16.48 16.67 16.45 16.475 69655425
2016-11-04 00:00:00+00:00 16.55 16.71 16.354 16.53 82431002
2016-11-07 00:00:00+00:00 17.01 17.04 16.85 16.86 91581902
2016-11-08 00:00:00+00:00 17 17.105 16.71 16.82 95702452
2016-11-09 00:00:00+00:00 17.97 18.05 17.4 17.66 319516881
2016-11-10 00:00:00+00:00 18.76 18.99 18.25 18.26 304889621
2016-11-11 00:00:00+00:00 19.02 19.03 18.63 18.64 212952487
2016-11-14 00:00:00+00:00 20.08 20.2 19.4 19.41 320959885
2016-11-15 00:00:00+00:00 20.16 20.18 19.6 19.79 190293076
2016-11-16 00:00:00+00:00 19.75 19.96 19.68 19.78 126662472
2016-11-17 00:00:00+00:00 20.08 20.22 19.705 19.76 167895828
2016-11-18 00:00:00+00:00 20 20.14 19.87 20.06 120620238
2016-11-21 00:00:00+00:00 20.33 20.35 20.01 20.1 117479106
2016-11-22 00:00:00+00:00 20.3 20.47 20.12 20.45 104720899
2016-11-23 00:00:00+00:00 20.56 20.66 20.2 20.576 110936499
2016-11-25 00:00:00+00:00 20.86 20.875 20.5 20.62 62610385
2016-11-28 00:00:00+00:00 20.3 20.85 20.25 20.67 108651231
2016-11-29 00:00:00+00:00 20.29 20.54 20.25 20.38 114159274
2016-11-30 00:00:00+00:00 21.12 21.19 20.72 20.77 191597643
2016-12-01 00:00:00+00:00 21.5 21.94 21.34 21.42 199092169
2016-12-02 00:00:00+00:00 21.23 21.5 21.02 21.49 128447563
2016-12-05 00:00:00+00:00 21.84 21.869 21.46 21.47 130377307
2016-12-06 00:00:00+00:00 22.16 22.16 21.72 22.09 120789194
2016-12-07 00:00:00+00:00 22.57 22.57 21.95 22.19 168005152
2016-12-08 00:00:00+00:00 22.95 23.24 22.69 22.79 182492861
2016-12-09 00:00:00+00:00 23.09 23.17 22.78 22.95 130590406
2016-12-12 00:00:00+00:00 22.61 23.25 22.535 23 136869792
2016-12-13 00:00:00+00:00 22.61 22.88 22.29 22.8 118204085
2016-12-14 00:00:00+00:00 22.67 23.3 22.21 22.31 225253082
2016-12-15 00:00:00+00:00 23.16 23.39 22.8 22.9 162025814
2016-12-16 00:00:00+00:00 22.66 23.32 22.65 23.32 147447864
2016-12-19 00:00:00+00:00 22.48 22.72 22.33 22.6 97439884
2016-12-20 00:00:00+00:00 22.71 22.83 22.61 22.63 83879520
2016-12-21 00:00:00+00:00 22.63 22.72 22.47 22.72 61046690
2016-12-22 00:00:00+00:00 22.54 22.73 22.47 22.6 67089135
2016-12-23 00:00:00+00:00 22.6 22.65 22.43 22.51 38172533
2016-12-27 00:00:00+00:00 22.61 22.735 22.54 22.71 39968380
2016-12-28 00:00:00+00:00 22.33 22.67 22.26 22.62 52619447
2016-12-29 00:00:00+00:00 22 22.39 21.77 22.33 79188397
2016-12-30 00:00:00+00:00 22.1 22.26 21.95 22.02 72592692
2017-01-03 00:00:00+00:00 22.53 22.68 22.2 22.6 99298080
2017-01-04 00:00:00+00:00 22.95 22.96 22.6 22.72 76846195
2017-01-05 00:00:00+00:00 22.68 22.93 22.345 22.82 86800397
2017-01-06 00:00:00+00:00 22.68 22.85 22.56 22.78 66281476
2017-01-09 00:00:00+00:00 22.55 22.71 22.4 22.51 75886389
2017-01-10 00:00:00+00:00 22.94 23.14 22.54 22.59 100977665
2017-01-11 00:00:00+00:00 23.07 23.07 22.72 22.94 92385551
2017-01-12 00:00:00+00:00 22.92 23.12 22.61 23.01 120274108
2017-01-13 00:00:00+00:00 23.01 23.409 22.8 23.21 161874990
2017-01-17 00:00:00+00:00 22.05 22.79 22.01 22.68 152495923
2017-01-18 00:00:00+00:00 22.63 22.65 22.1 22.3 124323059
2017-01-19 00:00:00+00:00 22.53 22.81 22.41 22.73 75990836
2017-01-20 00:00:00+00:00 22.64 22.93 22.52 22.66 102546886
2017-01-23 00:00:00+00:00 22.56 22.76 22.415 22.62 61333028
2017-01-24 00:00:00+00:00 22.95 23.095 22.48 22.61 98508678
2017-01-25 00:00:00+00:00 23.37 23.42 23.1 23.32 99753317
2017-01-26 00:00:00+00:00 23.44 23.55 23.28 23.41 84103638
2017-01-27 00:00:00+00:00 23.36 23.45 23.28 23.43 54558089
2017-01-30 00:00:00+00:00 22.95 23.2 22.71 23.2 91521352
2017-01-31 00:00:00+00:00 22.64 23.025 22.5 22.77 90978767
2017-02-01 00:00:00+00:00 22.89 23.22 22.82 22.97 103570051
2017-02-02 00:00:00+00:00 22.72 22.79 22.51 22.74 88679063
2017-02-03 00:00:00+00:00 23.29 23.35 22.95 23.15 115977015
2017-02-06 00:00:00+00:00 23.12 23.38 23.07 23.15 92207765
2017-02-07 00:00:00+00:00 22.9 23.29 22.86 23.28 87930117
2017-02-08 00:00:00+00:00 22.67 22.73 22.45 22.73 102250692
2017-02-09 00:00:00+00:00 23.12 23.15 22.64 22.76 102634174
2017-02-10 00:00:00+00:00 23.08 23.24 22.96 23.19 90548254
2017-02-13 00:00:00+00:00 23.4 23.54 23.17 23.17 105342461
2017-02-14 00:00:00+00:00 24.06 24.17 23.33 23.4 139611821
2017-02-15 00:00:00+00:00 24.58 24.77 24.11 24.34 151233334
2017-02-16 00:00:00+00:00 24.58 24.62 24.3 24.54 98144469
2017-02-17 00:00:00+00:00 24.52 24.58 24.2 24.28 85766765
2017-02-21 00:00:00+00:00 24.78 24.8 24.58 24.59 78522523
2017-02-22 00:00:00+00:00 24.79 24.95 24.54 24.61 81531895
2017-02-23 00:00:00+00:00 24.58 24.89 24.51 24.79 85845982
2017-02-24 00:00:00+00:00 24.23 24.35 24.02 24.12 97055343
2017-02-27 00:00:00+00:00 24.57 24.66 24.19 24.2 69057612
2017-02-28 00:00:00+00:00 24.68 24.7 24.42 24.48 90084039
2017-03-01 00:00:00+00:00 25.5 25.61 25.22 25.37 143947510
2017-03-02 00:00:00+00:00 25.23 25.8 25.2 25.68 99744502
2017-03-03 00:00:00+00:00 25.44 25.65 25.3 25.3 92681401
2017-03-06 00:00:00+00:00 25.25 25.35 25.08 25.33 75660430
2017-03-07 00:00:00+00:00 25.21 25.36 25.1 25.22 64019227
2017-03-08 00:00:00+00:00 25.26 25.77 25.22 25.6 105314569
2017-03-09 00:00:00+00:00 25.35 25.53 25.23 25.35 78984607
2017-03-10 00:00:00+00:00 25.31 25.62 25.09 25.62 86937037
2017-03-13 00:00:00+00:00 25.3 25.41 25.13 25.3 56886181
2017-03-14 00:00:00+00:00 25.32 25.34 25.05 25.19 63182745
2017-03-15 00:00:00+00:00 25.18 25.55 24.96 25.39 114662671
2017-03-16 00:00:00+00:00 25.22 25.49 25.19 25.24 70306233
2017-03-17 00:00:00+00:00 24.86 25.27 24.83 25.22 105222387
2017-03-20 00:00:00+00:00 24.44 24.76 24.42 24.58 87287477
2017-03-21 00:00:00+00:00 23.02 24.56 22.95 24.5 259545810
2017-03-22 00:00:00+00:00 22.94 23.18 22.45 22.65 167984438
2017-03-23 00:00:00+00:00 23.07 23.49 22.91 23.04 111666799
2017-03-24 00:00:00+00:00 23.12 23.35 22.76 23.2 113390157
2017-03-27 00:00:00+00:00 23.03 23.05 22.16 22.28 119513475
2017-03-28 00:00:00+00:00 23.48 23.67 23 23.01 95632231
2017-03-29 00:00:00+00:00 23.35 23.65 23.31 23.51 67204492
2017-03-30 00:00:00+00:00 23.87 23.98 23.33 23.38 79189235
2017-03-31 00:00:00+00:00 23.59 23.97 23.59 23.83 77158276
2017-04-03 00:00:00+00:00 23.59 23.75 23.08 23.65 87981757
2017-04-04 00:00:00+00:00 23.44 23.51 23.2 23.2 75219535
2017-04-05 00:00:00+00:00 23.17 23.88 23.15 23.77 97453336
2017-04-06 00:00:00+00:00 23.26 23.46 22.96 23.17 82975196
2017-04-07 00:00:00+00:00 23.16 23.34 22.93 23.03 79497011
2017-04-10 00:00:00+00:00 23.02 23.28 22.91 23.13 63254657
2017-04-11 00:00:00+00:00 22.92 22.95 22.58 22.83 94192834
2017-04-12 00:00:00+00:00 22.65 22.94 22.61 22.88 76556334
2017-04-13 00:00:00+00:00 22.34 22.96 22.34 22.56 88061760
2017-04-17 00:00:00+00:00 22.81 22.83 22.26 22.36 85053244
2017-04-18 00:00:00+00:00 22.71 23.15 22.38 23.11 146488440
2017-04-19 00:00:00+00:00 22.74 23.15 22.68 22.92 101771793
2017-04-20 00:00:00+00:00 23.07 23.17 22.82 22.96 103706737
2017-04-21 00:00:00+00:00 22.71 23.08 22.59 23.03 127235359
2017-04-24 00:00:00+00:00 23.63 23.85 23.24 23.24 138638933
2017-04-25 00:00:00+00:00 23.98 24.35 23.91 23.99 131531125
2017-04-26 00:00:00+00:00 23.89 24.19 23.86 24 95172221
2017-04-27 00:00:00+00:00 23.65 23.925 23.46 23.9 80024211
2017-04-28 00:00:00+00:00 23.34 23.72 23.32 23.57 69384908
2017-05-01 00:00:00+00:00 23.61 23.77 23.35 23.52 72382852
2017-05-02 00:00:00+00:00 23.53 23.67 23.33 23.61 60262632
2017-05-03 00:00:00+00:00 23.77 23.78 23.34 23.37 64179991
2017-05-04 00:00:00+00:00 23.85 24.05 23.73 23.98 68113987
2017-05-05 00:00:00+00:00 23.74 23.99 23.6 23.98 56678794
2017-05-08 00:00:00+00:00 23.96 24.06 23.75 23.75 56016460
2017-05-09 00:00:00+00:00 23.98 24.31 23.87 24.03 61621953
2017-05-10 00:00:00+00:00 24.15 24.17 23.87 23.91 48495255
2017-05-11 00:00:00+00:00 24.07 24.23 23.84 24.095 68350823
2017-05-12 00:00:00+00:00 24 24 23.75 23.92 62589008
2017-05-15 00:00:00+00:00 24.06 24.14 23.98 24.06 50517102
2017-05-16 00:00:00+00:00 23.99 24.17 23.94 24.07 55774774
2017-05-17 00:00:00+00:00 22.57 23.45 22.46 23.37 190353263
2017-05-18 00:00:00+00:00 22.74 23 22.5 22.61 108444096
2017-05-19 00:00:00+00:00 23.05 23.3 22.87 22.91 89566747
2017-05-22 00:00:00+00:00 23.04 23.24 22.82 23.19 74735545
2017-05-23 00:00:00+00:00 23.39 23.5 22.84 23.03 79020023
2017-05-24 00:00:00+00:00 23.36 23.47 23.22 23.44 59198571
2017-05-25 00:00:00+00:00 23.25 23.58 23.22 23.41 60435403
2017-05-26 00:00:00+00:00 23.24 23.42 23.17 23.23 50399458
2017-05-30 00:00:00+00:00 22.91 23.21 22.88 23.13 61307936
2017-05-31 00:00:00+00:00 22.41 22.8 22.09 22.78 138636343
2017-06-01 00:00:00+00:00 22.63 22.64 22.3 22.48 69629209
2017-06-02 00:00:00+00:00 22.45 22.59 22.16 22.27 79689080
2017-06-05 00:00:00+00:00 22.41 22.7 22.39 22.44 51590269
2017-06-06 00:00:00+00:00 22.23 22.41 22.07 22.2 71735574
2017-06-07 00:00:00+00:00 22.6 22.62 22.265 22.3 66366724
2017-06-08 00:00:00+00:00 22.97 23.21 22.6 22.62 93685867
2017-06-09 00:00:00+00:00 23.67 23.67 23.21 23.22 108254319
2017-06-12 00:00:00+00:00 23.78 24.035 23.55 23.82 85810394
2017-06-13 00:00:00+00:00 23.77 24.11 23.664 23.95 68047282
2017-06-14 00:00:00+00:00 23.76 23.81 23.2 23.55 102105009
2017-06-15 00:00:00+00:00 23.54 23.945 23.47 23.55 69707531
2017-06-16 00:00:00+00:00 23.43 23.62 23.37 23.56 57745812
2017-06-19 00:00:00+00:00 23.91 24 23.6 23.61 58544184
2017-06-20 00:00:00+00:00 23.49 23.99 23.48 23.91 59693544
2017-06-21 00:00:00+00:00 23.13 23.57 23.07 23.49 72432788
2017-06-22 00:00:00+00:00 22.93 23.195 22.9 23.12 64763274
2017-06-23 00:00:00+00:00 22.82 23.11 22.74 23.11 85399612
2017-06-26 00:00:00+00:00 22.89 23.14 22.73 22.91 61307843
2017-06-27 00:00:00+00:00 23.27 23.6 23 23.1 88257752
2017-06-28 00:00:00+00:00 23.88 23.96 23.529 23.53 118743321
2017-06-29 00:00:00+00:00 24.32 24.67 24.035 24.58 150501245
2017-06-30 00:00:00+00:00 24.26 24.62 24.17 24.62 83988074
2017-07-03 00:00:00+00:00 24.68 24.91 24.44 24.46 58119029
2017-07-05 00:00:00+00:00 24.92 24.95 24.62 24.8 76172541
2017-07-06 00:00:00+00:00 24.71 25.11 24.69 24.9 84465444
2017-07-07 00:00:00+00:00 24.83 25 24.65 24.92 54478781
2017-07-10 00:00:00+00:00 24.89 24.99 24.64 24.72 54632844
2017-07-11 00:00:00+00:00 24.6 24.91 24.46 24.88 69349320
2017-07-12 00:00:00+00:00 24.35 24.6 24.3 24.42 68708652
2017-07-13 00:00:00+00:00 24.62 24.63 24.37 24.43 55533384
2017-07-14 00:00:00+00:00 24.21 24.32 23.82 24.2 92554236
2017-07-17 00:00:00+00:00 24.02 24.22 23.93 24.22 81367339
2017-07-18 00:00:00+00:00 23.9 24.06 23.61 23.91 107981434
2017-07-19 00:00:00+00:00 24.06 24.2 23.81 24.05 74140055
2017-07-20 00:00:00+00:00 23.94 24.116 23.83 24.02 62775151
2017-07-21 00:00:00+00:00 23.8 23.96 23.66 23.84 65922474
2017-07-24 00:00:00+00:00 23.91 24.03 23.725 23.74 51366867
2017-07-25 00:00:00+00:00 24.48 24.67 24.24 24.25 86154719
2017-07-26 00:00:00+00:00 24.21 24.67 24.135 24.67 65704665
2017-07-27 00:00:00+00:00 24.11 24.45 24.034 24.26 62902110
2017-07-28 00:00:00+00:00 24.03 24.21 23.96 24.05 50143443
2017-07-31 00:00:00+00:00 24.12 24.3 24.09 24.12 61618916
2017-08-01 00:00:00+00:00 24.45 24.49 24.27 24.29 52782495
2017-08-02 00:00:00+00:00 24.59 24.59 24.26 24.44 55317130
2017-08-03 00:00:00+00:00 24.37 24.6 24.32 24.55 53168561
2017-08-04 00:00:00+00:00 24.97 25.08 24.65 24.68 105206610
2017-08-07 00:00:00+00:00 24.96 25.05 24.9 25.01 41860914
2017-08-08 00:00:00+00:00 24.9 25.35 24.82 24.95 73399327
2017-08-09 00:00:00+00:00 24.74 24.77 24.46 24.6 56910375
2017-08-10 00:00:00+00:00 24.12 24.58 24.09 24.58 77973831
2017-08-11 00:00:00+00:00 23.86 24.24 23.8 24.07 68157429
2017-08-14 00:00:00+00:00 24.42 24.485 24.17 24.23 54989495
2017-08-15 00:00:00+00:00 24.47 24.85 24.45 24.75 45143228
2017-08-16 00:00:00+00:00 24.19 24.62 24.08 24.54 59163774
2017-08-17 00:00:00+00:00 23.64 24.166 23.62 24.07 78969326
2017-08-18 00:00:00+00:00 23.62 23.9 23.45 23.56 60461197
2017-08-21 00:00:00+00:00 23.38 23.64 23.14 23.58 70106629
2017-08-22 00:00:00+00:00 23.83 23.89 23.57 23.6 59197032
2017-08-23 00:00:00+00:00 23.76 23.97 23.59 23.61 39825518
2017-08-24 00:00:00+00:00 23.84 23.94 23.68 23.93 39131435
2017-08-25 00:00:00+00:00 23.77 24.07 23.75 23.89 43511878
2017-08-28 00:00:00+00:00 23.72 23.9 23.57 23.84 37062661
2017-08-29 00:00:00+00:00 23.58 23.7 23.12 23.22 60846381
2017-08-30 00:00:00+00:00 23.87 24.04 23.56 23.73 58412886
2017-08-31 00:00:00+00:00 23.89 24.04 23.78 24.03 62362421
2017-09-01 00:00:00+00:00 24.09 24.18 23.8 23.9 56233871
2017-09-05 00:00:00+00:00 23.31 23.83 23.23 23.83 108981485
2017-09-06 00:00:00+00:00 23.41 23.575 23.21 23.49 64188863
2017-09-07 00:00:00+00:00 22.97 23.41 22.75 23.38 103471888
2017-09-08 00:00:00+00:00 22.89 23.23 22.85 22.93 68365012
2017-09-11 00:00:00+00:00 23.36 23.41 23.08 23.22 73741048
2017-09-12 00:00:00+00:00 23.95 24.03 23.45 23.46 99629843
2017-09-13 00:00:00+00:00 24.33 24.35 23.871 23.92 86339775
2017-09-14 00:00:00+00:00 24.24 24.54 24.2 24.38 69854207
2017-09-15 00:00:00+00:00 24.38 24.41 24.15 24.21 85937551
2017-09-18 00:00:00+00:00 24.7 24.77 24.41 24.42 68076369
2017-09-19 00:00:00+00:00 24.86 24.99 24.62 24.7 56190408
2017-09-20 00:00:00+00:00 25.06 25.24 24.67 24.88 82245970
2017-09-21 00:00:00+00:00 25.16 25.28 24.92 25.05 57838755
2017-09-22 00:00:00+00:00 25.02 25.07 24.85 25.01 61424966
2017-09-25 00:00:00+00:00 24.76 25.08 24.58 24.94 62309242
2017-09-26 00:00:00+00:00 24.81 24.955 24.685 24.83 52983481
2017-09-27 00:00:00+00:00 25.41 25.64 25.18 25.31 92809403
2017-09-28 00:00:00+00:00 25.45 25.64 25.34 25.59 64957669
2017-09-29 00:00:00+00:00 25.34 25.48 25.31 25.41 66211902
2017-10-02 00:00:00+00:00 25.62 25.65 25.39 25.46 54001378
2017-10-03 00:00:00+00:00 25.86 25.93 25.62 25.75 55913268
2017-10-04 00:00:00+00:00 25.71 25.95 25.7 25.86 53358291
2017-10-05 00:00:00+00:00 26.13 26.23 25.66 25.77 62636891
2017-10-06 00:00:00+00:00 26.21 26.3 26.04 26.25 53884276
2017-10-09 00:00:00+00:00 25.85 26.27 25.76 26.26 55153704
2017-10-10 00:00:00+00:00 25.93 25.95 25.71 25.83 46109501
2017-10-11 00:00:00+00:00 25.83 25.93 25.65 25.93 51639142
2017-10-12 00:00:00+00:00 25.45 25.93 25.34 25.87 71770031
2017-10-13 00:00:00+00:00 25.83 26 25.12 25.38 104143172
2017-10-16 00:00:00+00:00 26.24 26.33 25.87 25.87 71927178
2017-10-17 00:00:00+00:00 26.2 26.43 26.11 26.37 48589546
2017-10-18 00:00:00+00:00 26.48 26.55 26.33 26.34 55382456
2017-10-19 00:00:00+00:00 26.58 26.59 26.15 26.2 54493609
2017-10-20 00:00:00+00:00 27.17 27.18 26.9 27.04 83764655
2017-10-23 00:00:00+00:00 27.16 27.43 27.09 27.22 69578809
2017-10-24 00:00:00+00:00 27.68 27.84 27.34 27.35 90312306
2017-10-25 00:00:00+00:00 27.635 27.92 27.34 27.89 74516438
2017-10-26 00:00:00+00:00 27.74 27.98 27.68 27.69 55291119
2017-10-27 00:00:00+00:00 27.8 27.96 27.618 27.68 58856675
2017-10-30 00:00:00+00:00 27.6 27.76 27.46 27.65 51319946
2017-10-31 00:00:00+00:00 27.39 27.72 27.34 27.64 50311194
2017-11-01 00:00:00+00:00 27.53 27.73 27.36 27.64 46527703
2017-11-02 00:00:00+00:00 27.87 27.935 27.275 27.51 59447338
2017-11-03 00:00:00+00:00 27.82 27.82 27.62 27.73 37507075
2017-11-06 00:00:00+00:00 27.75 27.82 27.62 27.74 37009301
2017-11-07 00:00:00+00:00 27.18 27.77 27.02 27.74 67802318
2017-11-08 00:00:00+00:00 26.79 26.96 26.49 26.96 82724245
2017-11-09 00:00:00+00:00 26.49 26.66 26.12 26.49 94715530
2017-11-10 00:00:00+00:00 26.51 26.74 26.46 26.59 60715402
2017-11-13 00:00:00+00:00 26.4 26.48 26.14 26.26 55758728
2017-11-14 00:00:00+00:00 26.24 26.35 26.08 26.26 61283437
2017-11-15 00:00:00+00:00 26.79 26.93 25.81 25.99 98124987
2017-11-16 00:00:00+00:00 26.76 27.05 26.74 26.91 54329575
2017-11-17 00:00:00+00:00 26.62 26.73 26.445 26.51 48300548
2017-11-20 00:00:00+00:00 26.74 26.84 26.54 26.74 45279556
2017-11-21 00:00:00+00:00 26.73 26.86 26.69 26.8 57667414
2017-11-22 00:00:00+00:00 26.66 26.9 26.59 26.77 37984951
2017-11-24 00:00:00+00:00 26.59 26.74 26.57 26.72 14269207
2017-11-27 00:00:00+00:00 26.59 26.92 26.51 26.52 44606588
2017-11-28 00:00:00+00:00 27.64 27.7 26.565 26.62 95409925
2017-11-29 00:00:00+00:00 28.28 28.49 27.84 28.02 118070904
2017-11-30 00:00:00+00:00 28.17 28.72 28.08 28.62 120546324
2017-12-01 00:00:00+00:00 28.1 28.51 27.5 28.25 133921854
2017-12-04 00:00:00+00:00 29.06 29.31 28.82 28.85 143880151
2017-12-05 00:00:00+00:00 28.93 29.3 28.85 29.27 91034474
2017-12-06 00:00:00+00:00 28.64 28.97 28.57 28.73 73902305
2017-12-07 00:00:00+00:00 28.78 28.97 28.37 28.45 80517375
2017-12-08 00:00:00+00:00 29.05 29.09 28.76 29.08 60830895
2017-12-11 00:00:00+00:00 28.94 29.17 28.9 29.05 53947667
2017-12-12 00:00:00+00:00 29.32 29.5 28.91 29.05 75809939
2017-12-13 00:00:00+00:00 28.84 29.44 28.83 29.31 78318859
2017-12-14 00:00:00+00:00 28.73 29.18 28.71 29.065 59891544
2017-12-15 00:00:00+00:00 29.04 29.18 28.74 28.905 119265023
2017-12-18 00:00:00+00:00 29.48 29.5 29.27 29.32 69139708
2017-12-19 00:00:00+00:00 29.45 29.64 29.37 29.625 65397545
2017-12-20 00:00:00+00:00 29.48 29.84 29.45 29.765 81212183
2017-12-21 00:00:00+00:00 29.82 29.98 29.58 29.595 72441696
2017-12-22 00:00:00+00:00 29.88 30.03 29.62 29.965 54073301
2017-12-26 00:00:00+00:00 29.78 29.94 29.58 29.74 42254735
2017-12-27 00:00:00+00:00 29.73 29.73 29.601 29.66 31827793
2017-12-28 00:00:00+00:00 29.8 29.82 29.66 29.73 37885624
2017-12-29 00:00:00+00:00 29.52 29.88 29.52 29.85 40133412
2018-01-02 00:00:00+00:00 29.9 29.9 29.61 29.75 56984205
2018-01-03 00:00:00+00:00 29.8 29.94 29.69 29.9 57770689
2018-01-04 00:00:00+00:00 30.19 30.44 29.88 29.97 75653541
2018-01-05 00:00:00+00:00 30.33 30.42 30.05 30.37 56124567
2018-01-08 00:00:00+00:00 30.12 30.27 30.05 30.23 42109249
2018-01-09 00:00:00+00:00 30.27 30.54 30.13 30.2 69281639
2018-01-10 00:00:00+00:00 30.55 30.73 30.31 30.37 62272175
2018-01-11 00:00:00+00:00 30.66 30.69 30.45 30.66 59015496
2018-01-12 00:00:00+00:00 31.19 31.2 30.77 30.88 66034385
2018-01-16 00:00:00+00:00 31.24 31.79 31.03 31.74 104217423
2018-01-17 00:00:00+00:00 31.18 31.29 30.34 31 122797447
2018-01-18 00:00:00+00:00 31.48 31.71 31.21 31.33 76276795
2018-01-19 00:00:00+00:00 31.72 31.74 31.46 31.58 66236420
2018-01-22 00:00:00+00:00 31.94 31.94 31.58 31.67 52627095
2018-01-23 00:00:00+00:00 31.92 32.13 31.785 31.86 56124675
2018-01-24 00:00:00+00:00 32.09 32.21 31.82 32.009 80706611
2018-01-25 00:00:00+00:00 32.09 32.25 31.93 32.24 60230741
2018-01-26 00:00:00+00:00 32.2 32.2 31.95 32.11 50334774
2018-01-29 00:00:00+00:00 32.28 32.45 32.18 32.25 58205136
2018-01-30 00:00:00+00:00 31.88 32.2 31.85 31.95 58917317
2018-01-31 00:00:00+00:00 32 32.29 31.95 32.049 65150630
2018-02-01 00:00:00+00:00 32.5 32.5 31.96 32 62054274
2018-02-02 00:00:00+00:00 31.95 32.67 31.86 32.439 96387301
2018-02-05 00:00:00+00:00 30.26 31.98 29.15 31.12 153479504
2018-02-06 00:00:00+00:00 31.2 31.29 29.3 29.41 161141551
2018-02-07 00:00:00+00:00 31.25 31.74 30.86 31.12 99831300
2018-02-08 00:00:00+00:00 29.74 31.38 29.73 31.31 128282047
2018-02-09 00:00:00+00:00 30.33 30.61 29.13 30.17 142693575
2018-02-12 00:00:00+00:00 31.12 31.44 30.54 30.62 94696662
2018-02-13 00:00:00+00:00 31.18 31.4 30.76 31.04 76559902
2018-02-14 00:00:00+00:00 32 32.03 31.055 31.17 92773022
2018-02-15 00:00:00+00:00 32.21 32.47 31.87 32.39 67806215
2018-02-16 00:00:00+00:00 31.97 32.259 31.92 32 61805677
2018-02-20 00:00:00+00:00 31.93 32.159 31.75 31.96 58356720
2018-02-21 00:00:00+00:00 31.87 32.365 31.77 31.8 71980440
2018-02-22 00:00:00+00:00 31.69 32.14 31.59 31.99 64394876
2018-02-23 00:00:00+00:00 32.03 32.06 31.76 31.8 50959564
2018-02-26 00:00:00+00:00 32.42 32.439 32.049 32.17 60441481
2018-02-27 00:00:00+00:00 32.33 32.85 32.32 32.32 68943159
2018-02-28 00:00:00+00:00 32.1 32.77 32.09 32.49 70351619
2018-03-01 00:00:00+00:00 31.48 32.345 31.36 32.07 90292066
2018-03-02 00:00:00+00:00 31.63 31.69 30.63 31.12 82426632
2018-03-05 00:00:00+00:00 32.13 32.354 31.165 31.35 70074106
2018-03-06 00:00:00+00:00 32.11 32.32 31.95 32.299 50302159
2018-03-07 00:00:00+00:00 32.18 32.229 31.58 31.7 52710046
2018-03-08 00:00:00+00:00 32.2 32.27 31.81 32.189 51206287
2018-03-09 00:00:00+00:00 32.72 32.729 32.4 32.47 73075006
2018-03-12 00:00:00+00:00 32.84 33.05 32.64 32.689 59301639
2018-03-13 00:00:00+00:00 32.36 32.99 32.299 32.97 62823477
2018-03-14 00:00:00+00:00 32.14 32.549 31.93 32.549 57889408
2018-03-15 00:00:00+00:00 32.1 32.34 31.93 32.29 41825422
2018-03-16 00:00:00+00:00 32.17 32.57 32.08 32.11 82176070
2018-03-19 00:00:00+00:00 31.98 32.22 31.68 32.13 57068585
2018-03-20 00:00:00+00:00 31.98 32.2 31.88 32.06 44455976
2018-03-21 00:00:00+00:00 31.87 32.45 31.68 32 64135559
2018-03-22 00:00:00+00:00 30.55 31.49 30.42 31.44 109386636
2018-03-23 00:00:00+00:00 29.17 30.82 29.05 30.69 114544472
2018-03-26 00:00:00+00:00 30.44 30.555 29.65 29.89 81989080
2018-03-27 00:00:00+00:00 29.52 30.67 29.21 30.65 85188716
2018-03-28 00:00:00+00:00 29.39 29.96 29.01 29.79 76177788
2018-03-29 00:00:00+00:00 29.99 30.14 29.31 29.57 65246291
2018-04-02 00:00:00+00:00 29.31 30.055 28.75 29.8 85943825
2018-04-03 00:00:00+00:00 29.59 29.65 29.1 29.54 69497275
2018-04-04 00:00:00+00:00 29.88 29.97 28.95 29 65831606
2018-04-05 00:00:00+00:00 30.32 30.55 30.115 30.18 55411276
2018-04-06 00:00:00+00:00 29.63 30.29 29.37 30.01 79474397
2018-04-09 00:00:00+00:00 29.87 30.625 29.76 29.82 65530400
2018-04-10 00:00:00+00:00 30.48 30.58 30.23 30.48 60540481
2018-04-11 00:00:00+00:00 29.9 30.24 29.89 30.15 61963905
2018-04-12 00:00:00+00:00 30.65 30.8 30.165 30.22 57562976
2018-04-13 00:00:00+00:00 29.8 31.17 29.56 31.13 98811401
2018-04-16 00:00:00+00:00 29.93 30.37 29.47 30.08 109100369
2018-04-17 00:00:00+00:00 30.04 30.37 29.88 30.17 82744702
2018-04-18 00:00:00+00:00 29.53 30.22 29.5 30.03 79135270
2018-04-19 00:00:00+00:00 30.18 30.24 29.54 29.55 80740174
2018-04-20 00:00:00+00:00 30.26 30.53 30.13 30.26 63123024
2018-04-23 00:00:00+00:00 30.32 30.4 30.12 30.27 50308425
2018-04-24 00:00:00+00:00 30.19 30.86 30 30.46 81153889
2018-04-25 00:00:00+00:00 30.14 30.34 29.8 30.09 64240382
2018-04-26 00:00:00+00:00 30.07 30.28 29.99 30.04 44862773
2018-04-27 00:00:00+00:00 30.15 30.21 29.97 29.99 43905512
2018-04-30 00:00:00+00:00 29.92 30.43 29.92 30.27 54818739
2018-05-01 00:00:00+00:00 29.95 29.96 29.62 29.92 52630875
2018-05-02 00:00:00+00:00 29.58 30.07 29.56 29.95 67879858
2018-05-03 00:00:00+00:00 29.2 29.52 28.43 29.51 110445605
2018-05-04 00:00:00+00:00 29.3 29.42 28.82 28.99 58166030
2018-05-07 00:00:00+00:00 29.64 29.788 29.33 29.44 46537019
2018-05-08 00:00:00+00:00 29.93 30.125 29.665 29.77 64647835
2018-05-09 00:00:00+00:00 30.72 30.79 30.08 30.1 71822529
2018-05-10 00:00:00+00:00 30.89 31.07 30.42 30.61 54475803
2018-05-11 00:00:00+00:00 30.92 31.03 30.81 30.91 41930557
2018-05-14 00:00:00+00:00 31.12 31.23 31.03 31.04 41634010
2018-05-15 00:00:00+00:00 31.22 31.36 30.85 30.92 62290774
2018-05-16 00:00:00+00:00 31.06 31.25 31.02 31.18 43337537
2018-05-17 00:00:00+00:00 30.81 31.08 30.67 31.04 45566657
2018-05-18 00:00:00+00:00 30.26 30.79 30.23 30.77 55842588
2018-05-21 00:00:00+00:00 30.55 30.68 30.48 30.52 34592438
2018-05-22 00:00:00+00:00 30.89 31.14 30.54 30.6 60720402
2018-05-23 00:00:00+00:00 30.44 30.76 30.081 30.71 73081366
2018-05-24 00:00:00+00:00 30.21 30.41 29.83 30.4 61592751
2018-05-25 00:00:00+00:00 30.16 30.19 29.88 30.01 43955277
2018-05-29 00:00:00+00:00 28.96 29.8 28.7 29.76 135380296
2018-05-30 00:00:00+00:00 29.49 29.6 29.16 29.29 78861008
2018-05-31 00:00:00+00:00 29.04 29.26 28.825 29.2 94612359
2018-06-01 00:00:00+00:00 29.4 29.66 29.31 29.49 64580957
2018-06-04 00:00:00+00:00 29.4 29.59 29.33 29.55 42386702
2018-06-05 00:00:00+00:00 29.12 29.34 29.06 29.31 46709649
2018-06-06 00:00:00+00:00 30.04 30.05 29.28 29.29 74299170
2018-06-07 00:00:00+00:00 30.09 30.31 29.89 30.2 66638273
2018-06-08 00:00:00+00:00 30.01 30.05 29.81 30.01 49186782
2018-06-11 00:00:00+00:00 30.06 30.41 30.06 30.13 46578051
2018-06-12 00:00:00+00:00 29.9 30.27 29.75 30.17 49210024
2018-06-13 00:00:00+00:00 29.84 30.355 29.695 29.98 68282762
2018-06-14 00:00:00+00:00 29.5 30.05 29.38 30.04 77025742
2018-06-15 00:00:00+00:00 29.28 29.41 28.865 29.25 102217511
2018-06-18 00:00:00+00:00 29.4 29.5 28.81 28.94 52299539
2018-06-19 00:00:00+00:00 29.26 29.37 28.95 29.04 62582970
2018-06-20 00:00:00+00:00 29.24 29.5 29.18 29.42 51983832
2018-06-21 00:00:00+00:00 29.29 29.45 29 29.21 67998508
2018-06-22 00:00:00+00:00 28.99 29.52 28.99 29.46 76079178
2018-06-25 00:00:00+00:00 28.48 28.91 28.26 28.86 72741314
2018-06-26 00:00:00+00:00 28.54 28.67 28.01 28.52 55889852
2018-06-27 00:00:00+00:00 28.24 28.845 28.24 28.51 61066628
2018-06-28 00:00:00+00:00 28.67 28.84 28.3 28.4 69893231
2018-06-29 00:00:00+00:00 28.19 29.17 28.18 29.09 101895880
2018-07-02 00:00:00+00:00 28.28 28.28 27.83 28.08 52090546
2018-07-03 00:00:00+00:00 27.78 28.43 27.74 28.33 40597682
2018-07-05 00:00:00+00:00 27.92 28.079 27.81 27.95 44123567
2018-07-06 00:00:00+00:00 28.03 28.15 27.63 27.81 39424933
2018-07-09 00:00:00+00:00 29.05 29.09 28.22 28.23 60593658
2018-07-10 00:00:00+00:00 28.83 29.33 28.75 29.22 58271064
2018-07-11 00:00:00+00:00 28.68 28.91 28.62 28.67 48936591
2018-07-12 00:00:00+00:00 28.77 28.97 28.66 28.9 50729700
2018-07-13 00:00:00+00:00 28.55 28.69 28.221 28.62 74861603
2018-07-16 00:00:00+00:00 29.78 29.85 28.73 28.78 128205773
2018-07-17 00:00:00+00:00 30.01 30.2 29.58 29.89 86967476
2018-07-18 00:00:00+00:00 30.13 30.29 29.84 29.92 58908069
2018-07-19 00:00:00+00:00 29.67 30.015 29.64 29.93 59123021
2018-07-20 00:00:00+00:00 30.13 30.16 29.46 29.62 80292243
2018-07-23 00:00:00+00:00 30.75 30.85 30.13 30.14 73723538
2018-07-24 00:00:00+00:00 30.83 31.11 30.65 30.85 64652784
2018-07-25 00:00:00+00:00 31.07 31.11 30.67 30.7 57391699
2018-07-26 00:00:00+00:00 30.94 31.21 30.89 31.19 45903175
2018-07-27 00:00:00+00:00 31.06 31.14 30.8 30.99 56979036
2018-07-30 00:00:00+00:00 31.31 31.45 31.13 31.14 58393904
2018-07-31 00:00:00+00:00 30.88 31.415 30.82 31.36 63625038
2018-08-01 00:00:00+00:00 31.25 31.55 31.17 31.22 68471546
2018-08-02 00:00:00+00:00 31.28 31.31 30.87 30.98 46003576
2018-08-03 00:00:00+00:00 31.51 31.51 31.16 31.18 49931310
2018-08-06 00:00:00+00:00 31.52 31.59 31.325 31.45 41079640
2018-08-07 00:00:00+00:00 31.51 31.8 31.49 31.55 43604958
2018-08-08 00:00:00+00:00 31.8 31.91 31.48 31.51 43288591
2018-08-09 00:00:00+00:00 31.6 31.79 31.55 31.73 37740761
2018-08-10 00:00:00+00:00 31.19 31.325 30.9 31.23 54712011
2018-08-13 00:00:00+00:00 30.48 31.12 30.45 31.08 62834822
2018-08-14 00:00:00+00:00 30.79 30.9 30.42 30.61 42172878
2018-08-15 00:00:00+00:00 30.36 30.62 30.155 30.48 59326967
2018-08-16 00:00:00+00:00 30.72 30.84 30.52 30.56 43583825
2018-08-17 00:00:00+00:00 30.74 30.855 30.545 30.66 38275432
2018-08-20 00:00:00+00:00 30.87 30.925 30.65 30.69 39889778
2018-08-21 00:00:00+00:00 31.02 31.25 30.83 30.91 53466881
2018-08-22 00:00:00+00:00 30.98 31.085 30.88 30.92 33000899
2018-08-23 00:00:00+00:00 30.84 30.98 30.74 30.94 37777195
2018-08-24 00:00:00+00:00 30.89 31.08 30.85 30.94 42964476
2018-08-27 00:00:00+00:00 31.31 31.49 30.99 31.03 47364148
2018-08-28 00:00:00+00:00 31.27 31.41 31.19 31.36 38447811
2018-08-29 00:00:00+00:00 31.14 31.31 31.07 31.24 45844152
2018-08-30 00:00:00+00:00 31.01 31.15 30.95 31.02 46559486
2018-08-31 00:00:00+00:00 30.93 30.95 30.62 30.88 48595082
2018-09-04 00:00:00+00:00 31.14 31.16 30.82 30.91 44527968
2018-09-05 00:00:00+00:00 31.18 31.24 30.97 31.07 46427457
2018-09-06 00:00:00+00:00 30.85 31.11 30.73 31.05 41692185
2018-09-07 00:00:00+00:00 30.86 31.12 30.63 31 48514924
2018-09-10 00:00:00+00:00 30.82 31 30.75 30.88 33065157
2018-09-11 00:00:00+00:00 30.85 30.9 30.7 30.72 48391250
2018-09-12 00:00:00+00:00 30.43 30.96 30.38 30.88 58933386
2018-09-13 00:00:00+00:00 30.14 30.67 30.08 30.53 57950233
2018-09-14 00:00:00+00:00 30.37 30.39 30.11 30.13 33103669
2018-09-17 00:00:00+00:00 30.28 30.49 30.15 30.34 34050278
2018-09-18 00:00:00+00:00 30.21 30.4 30.13 30.33 49837464
2018-09-19 00:00:00+00:00 31 31.2 30.13 30.13 91304426
2018-09-20 00:00:00+00:00 31.19 31.37 31.14 31.3 82751190
2018-09-21 00:00:00+00:00 31.03 31.37 30.97 31.34 85445286
2018-09-24 00:00:00+00:00 30.74 31.2 30.66 30.98 45353607
2018-09-25 00:00:00+00:00 30.67 31.04 30.6 30.85 40715873
2018-09-26 00:00:00+00:00 30.13 30.74 30.06 30.71 57222952
2018-09-27 00:00:00+00:00 29.94 30.31 29.93 30.24 45462288
2018-09-28 00:00:00+00:00 29.46 29.85 29.42 29.65 73275235
2018-10-01 00:00:00+00:00 29.65 29.94 29.54 29.68 48539096
2018-10-02 00:00:00+00:00 29.58 29.72 29.27 29.58 42786255
2018-10-03 00:00:00+00:00 30 30.175 29.72 29.81 60663616
2018-10-04 00:00:00+00:00 30.43 30.79 30.14 30.17 71995080
2018-10-05 00:00:00+00:00 30.23 30.65 30.05 30.6 51722674
2018-10-08 00:00:00+00:00 30.27 30.42 29.91 30.05 47054908
2018-10-09 00:00:00+00:00 29.98 30.19 29.87 30.03 57490810
2018-10-10 00:00:00+00:00 29.24 30.13 29.21 29.98 88338165
2018-10-11 00:00:00+00:00 28.36 29.25 28.26 28.89 111417275
2018-10-12 00:00:00+00:00 28.46 29.04 27.73 28.99 101947113
2018-10-15 00:00:00+00:00 27.92 28.62 27.64 28.49 112963592
2018-10-16 00:00:00+00:00 28.53 28.58 27.9 28.24 87633865
2018-10-17 00:00:00+00:00 28.9 29.19 28.16 28.37 78596745
2018-10-18 00:00:00+00:00 28.25 28.9 28.14 28.67 77080398
2018-10-19 00:00:00+00:00 28.32 28.52 27.97 28.14 64231096
2018-10-22 00:00:00+00:00 27.38 28.35 27.31 28.32 89660977
2018-10-23 00:00:00+00:00 27.02 27.2 26.105 26.39 102244478
2018-10-24 00:00:00+00:00 26.19 26.9 26.08 26.86 85041168
2018-10-25 00:00:00+00:00 26.59 26.93 26.18 26.38 73309321
2018-10-26 00:00:00+00:00 26.39 26.58 25.88 26.33 90904719
2018-10-29 00:00:00+00:00 26.61 27.18 26.29 26.69 87404322
2018-10-30 00:00:00+00:00 26.78 26.86 26.32 26.76 78716947
2018-10-31 00:00:00+00:00 27.5 27.88 27.06 27.1 91899324
2018-11-01 00:00:00+00:00 27.81 28 27.61 27.77 49121761
2018-11-02 00:00:00+00:00 27.89 28.38 27.52 28.01 68878579
2018-11-05 00:00:00+00:00 28.06 28.24 27.89 27.95 40329161
2018-11-06 00:00:00+00:00 28.21 28.245 27.74 28.02 43143905
2018-11-07 00:00:00+00:00 28.54 28.67 28.14 28.415 55645361
2018-11-08 00:00:00+00:00 28.87 29.12 28.43 28.5 48930851
2018-11-09 00:00:00+00:00 28.52 28.89 28.384 28.75 43102533
2018-11-12 00:00:00+00:00 27.75 28.52 27.64 28.38 50160927
2018-11-13 00:00:00+00:00 27.76 28.14 27.69 27.75 57304289
2018-11-14 00:00:00+00:00 27.21 28.1 26.78 27.95 68623979
2018-11-15 00:00:00+00:00 27.9 27.97 26.88 27.14 65582163
2018-11-16 00:00:00+00:00 27.75 27.89 27.42 27.69 52223527
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_nn_Challenges in Live Trading_Execution Model-NN py3.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_nn_Challenges in Live Trading_Keras_CV.py
from keras.models import Sequential
from keras.layers import Dropout
from keras.layers import Dense, Activation
def create_new_model(neurons=20,act_1='sigmoid',dropout_ratio=0.15):
model=Sequential()
model.add(Dense(neurons,
kernel_initializer='he_normal'
,input_shape=(5,)
, bias_initializer='zeros'))
model.add(Activation(act_1))
model.add(Dropout(dropout_ratio))
model.add(Dense(neurons*2, use_bias=True, kernel_initializer='he_normal'
, bias_initializer='zeros'))
model.add(Activation(act_1))
model.add(Dropout(dropout_ratio))
model.add(Dense(neurons*3, use_bias=True, kernel_initializer='he_normal'
, bias_initializer='zeros'))
model.add(Activation(act_1))
model.add(Dropout(dropout_ratio))
model.add(Dense(neurons*4, use_bias=True, kernel_initializer='he_normal'
, bias_initializer='zeros'))
model.add(Activation(act_1))
model.add(Dropout(dropout_ratio))
model.add(Dense(neurons*5, use_bias=True, kernel_initializer='he_normal'
, bias_initializer='zeros'))
model.add(Activation(act_1))
model.add(Dropout(dropout_ratio))
model.add(Dense(1,activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics = ['accuracy'])
return model
Raw
logicbelief-master-puwork_nn_Challenges in Live Trading_model_save.json
"{\"class_name\": \"Sequential\", \"config\": {\"name\": \"sequential_46\", \"layers\": [{\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_271\", \"trainable\": true, \"batch_input_shape\": [null, 5], \"dtype\": \"float32\", \"units\": 8, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"normal\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"Activation\", \"config\": {\"name\": \"activation_226\", \"trainable\": true, \"activation\": \"relu\"}}, {\"class_name\": \"Dropout\", \"config\": {\"name\": \"dropout_226\", \"trainable\": true, \"rate\": 0.02, \"noise_shape\": null, \"seed\": null}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_272\", \"trainable\": true, \"units\": 16, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"normal\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"Activation\", \"config\": {\"name\": \"activation_227\", \"trainable\": true, \"activation\": \"relu\"}}, {\"class_name\": \"Dropout\", \"config\": {\"name\": \"dropout_227\", \"trainable\": true, \"rate\": 0.02, \"noise_shape\": null, \"seed\": null}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_273\", \"trainable\": true, \"units\": 24, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"normal\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"Activation\", \"config\": {\"name\": \"activation_228\", \"trainable\": true, \"activation\": \"relu\"}}, {\"class_name\": \"Dropout\", \"config\": {\"name\": \"dropout_228\", \"trainable\": true, \"rate\": 0.02, \"noise_shape\": null, \"seed\": null}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_274\", \"trainable\": true, \"units\": 32, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"normal\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"Activation\", \"config\": {\"name\": \"activation_229\", \"trainable\": true, \"activation\": \"relu\"}}, {\"class_name\": \"Dropout\", \"config\": {\"name\": \"dropout_229\", \"trainable\": true, \"rate\": 0.02, \"noise_shape\": null, \"seed\": null}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_275\", \"trainable\": true, \"units\": 40, \"activation\": \"linear\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 2.0, \"mode\": \"fan_in\", \"distribution\": \"normal\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}, {\"class_name\": \"Activation\", \"config\": {\"name\": \"activation_230\", \"trainable\": true, \"activation\": \"relu\"}}, {\"class_name\": \"Dropout\", \"config\": {\"name\": \"dropout_230\", \"trainable\": true, \"rate\": 0.02, \"noise_shape\": null, \"seed\": null}}, {\"class_name\": \"Dense\", \"config\": {\"name\": \"dense_276\", \"trainable\": true, \"units\": 1, \"activation\": \"sigmoid\", \"use_bias\": true, \"kernel_initializer\": {\"class_name\": \"VarianceScaling\", \"config\": {\"scale\": 1.0, \"mode\": \"fan_avg\", \"distribution\": \"uniform\", \"seed\": null}}, \"bias_initializer\": {\"class_name\": \"Zeros\", \"config\": {}}, \"kernel_regularizer\": null, \"bias_regularizer\": null, \"activity_regularizer\": null, \"kernel_constraint\": null, \"bias_constraint\": null}}]}, \"keras_version\": \"2.2.4\", \"backend\": \"tensorflow\"}"
Raw
logicbelief-master-puwork_README.md

逻辑信仰

目标

建立量化的投资系统， 包括以下模块

  1. 量化选股
  2. 回测
  3. 策略开发
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_nn_Cross Validation in Keras_Cross Validation in Keras-Upload py3.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_nn_Cross Validation in Keras_Keras_CV.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_nn_Cross Validation in Keras_SBIN.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_nn_Deep Learning in Trading_Deep Neural Network Prediction Model for trading- Upload.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_nn_Deep Learning in Trading_model.json
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_nn_Deep Learning in Trading_SBIN.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_nn_Long Short Term Memory Unit(LSTMs)_LSTM- Price Prediction-Upload.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_nn_Long Short Term Memory Unit(LSTMs)_SBIN.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_nn_Neural Networks_BA.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_nn_Neural Networks_Neural Network.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_nn_Recurrent Neural Networks_RNN- Example.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_nn_Recurrent Neural Networks_SBIN.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_nn_sample_NN_deploy_strategy.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_100.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_1000.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_3p.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_3p_reselection.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_500.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_5year_daydata_cvs.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_backtester.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_basic_eps.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_best.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_breakout_test.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_notebook_D__a.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_daily.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_daily_basic.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_days200.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_demo.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_down_days_bt.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_finance.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_golden.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_hour.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_hurst.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_ibd50.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_ma_bt.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_ma_diff_bt.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_mark.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_mark_bt.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_master.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_new_high_low.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_position_watcher.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_positive_eps.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_prime.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_pullback.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_realtime.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_selection.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_selection_backtest.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_single.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_stock_find.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_t_zi_ban.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_view.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_vma60.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_winner.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_winner_vs_loser_finance.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_yi_zi_ban.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_zhang_ting.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_notebook_zhangting_backtest.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_one_concept_buy.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_one_industry_buy.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_predict_local.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_predict_mark.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_predict_sepa.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_profit.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_profit_test.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_quantos_data.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_quantos_demoStrategy_configs.json
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_quantos_demoStrategy_dataview_meta_data.json
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_quantos_demoStrategy_trades.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_quantos_double_ma_configs.json
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_quantos_double_ma_report.html
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_quantos_double_ma_trades.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture10_homework_future_cash_basis.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture12_DoubleMAStrats.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture13_CalendarSpread.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_quantos_lecture16_industry_daily_ret.csv
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture22_BacktestExample.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture2_DemoStrategy.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture3_medianRet.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture4_candelstick.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_lecture9_position_management.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_Live20180206.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Display the source blob
Display the rendered blob
Raw
logicbelief-master-puwork_quantos_tk_technical_analysis.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Raw
logicbelief-master-puwork_realtime_buy.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_realtime_close_buy.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_realtime_concept.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_realtime_sell.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_reverse_buy.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_reverse_buy_bt.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Raw
logicbelief-master-puwork_reverse_hk_local.py
View raw

(Sorry about that, but we can’t show files that are this big right now.)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment