marufeuille

def generate(phi_1, phi_2, c, sigma, size):
    # 期待値
    mu = c/(1-phi_1-phi_2)

    # 乱数のseedを固定
    np.random.seed(17)

    # データの生成
    ar_data = np.zeros(size)
    ar_data[0] = mu + np.random.normal(0, sigma)

marufeuille

from aliyun.log import *
from aliyun.log.util import base64_encodestring

# アップロード用の関数を定義
def put_logs(client, project, logstore, contents, compress=False):
    topic = ''
    source = ''
    logitemList = []  # LogItem list
    for c in contents:
        logItem = LogItem()

marufeuille

# データの生成
ar_data = generate(1.4, -0.48, 5, 0.5, 100)

# データをアップロードしやすい形式に変換
timestamps = sorted(range(0, 100))
contents = []
for timestamp in timestamps:
    contents.append([
        ("timestamp", str(timestamp)), # 必ず文字列に変換すること
        ("value", str(ar_data[timestamp]))

marufeuille

* | select
    ts_predicate_ar(timestamp, value, 2, 10)
limit 110

marufeuille

* | select
    ts_cp_detect(timestamp, value, 50)
limit 1000

marufeuille

* | select
    unixtime,
    predict
from (
    select
        ts_predicate_ar(timestamp, value, 2, 10)
    from log
)

marufeuille

* | select
    preds[1] as unixtime,
    preds[3] as predict
from (
    select
        ts_predicate_ar(timestamp, value, 2, 10) as p
    from log
), unnest(p) as t(preds)

marufeuille

select
    timestamp,
    value
where
    timestamp > ( to_unixtime(localtimestamp) - 3600)
order by timestamp asc

marufeuille

ar_data = np.concatenate((
    generate(1.4, -0.48, 5, 0.5, 450),
    generate(1.4, -0.48, 10, 0.5, 100),
    generate(1.4, -0.48, 5, 0.5, 450)
))
timestamps = sorted(range(1000))
contents = []
for timestamp in timestamps:
    contents.append([
        ("timestamp", str(timestamp)),

marufeuille

data = pd.read_csv("./day.csv")
for i in range(data.shape[0]):
    contents = [list(zip(data.columns, map(lambda x: str(x), tuple(data.loc[i,:]))))]
    put_logs(client, project, logstore, contents)