株価間倍率を計算してグラフ化します。

arbitrage.py

from datetime import datetime as dt
import datetime
import pandas as pd
from pandas import DataFrame
import csv
from functools import reduce
import math
import matplotlib
import matplotlib.pyplot as plt

matplotlib.style.use('ggplot')

date = dt.now()
datestr = date.strftime('%Y%m%d')

def calc():
    list_file_name = 'data/stock_list_%s.csv' % datestr
    stock_list1 = pd.read_csv(list_file_name, encoding = 'shift-jis')
    stock_list2 = pd.read_csv(list_file_name, encoding = 'shift-jis')
    for k1, stock1 in stock_list1.iterrows():
        code1 = stock1['コード']
        try:
            df1 = code2df(code1)
        except:
            continue
        var_list = {}
        df_all = DataFrame(columns=(code1,))
        for k2, stock2 in stock_list2.iterrows():
            code2 = stock2['コード']
            if code1 == code2:
                continue
            try:
                df2 = code2df(code2)
            except:
                continue
            l = []
            df_ratio = DataFrame(columns=(code2,))
            for i, row1 in df1.iterrows():
                try:
                    row2 =  df2.loc[df2['日付'] == row1['日付']]
                    ratio = float(row1['終値']) / float(row2['終値'])
                    df_ratio.loc[row1['日付']] = [ratio]
                    l.append(ratio)
                except:
                    #print(row1['終値'], row2['終値'])
                    continue
            # filter nan
            l = list(filter(lambda v: v==v, l))
            avg = float(sum(l)) / len(l)
            # calculate standard diviation / average
            sd = math.sqrt(sum(map(lambda x: pow((x - avg), 2), l))  / len(l))
            var = sd / avg
            last_diff = abs(l[-1] - avg)
            print(code1, code2, 'avg:', avg, 'sd:', sd, 'var:', var, 'last_diff:', last_diff)
            if math.isnan(sd) == False:
                var_list[code2] = var
                df_all = pd.concat([df_all, df_ratio], axis=1)

        # plot variation and save image
        sorted_var = sorted(var_list.items(), key=lambda x: x[1])
        for k,v in sorted_var:
          if code1 < k:
            print(code1,k,v)
            df = df_all[k]
            plot = df.plot()
            fig = plot.get_figure()
            fig.savefig("img/%s_%s.png" % (code1, k))
            fig.clf()

def get_file_name(code, datestr):
    return "data/%s_%s.csv" % (code, datestr)

def code2df(code):
    filename = get_file_name(code, datestr)
    return pd.read_csv(filename, encoding = 'shift-jis')

if __name__ == "__main__":
    calc()