TSAP : TimeSeries Analysis with Python

%matplotlib inline 
import matplotlib.pylab
import numpy as np
import pandas as pd

Window functions are like aggregation functions

df = pd.DataFrame(np.random.randn(300, 3), 
                  index = pd.date_range('7/1/2016', freq = 'D',periods = 300),
                  columns = ['A', 'B', 'C'])

df.head()

	A	B	C
2016-07-01	1.251741	-0.669649	0.845530
2016-07-02	-1.065566	0.399674	-0.034806
2016-07-03	-0.990938	-0.214281	-1.890150
2016-07-04	2.084868	1.041457	0.924540
2016-07-05	1.762719	-1.435734	-1.234330

#pd.rolling_mean(df, window = 2)[1:10] # in future versions you want to resample separately
r = df.rolling(window = 50)
#r.agg, r.apply, r.count, r.exclusions,
#r.max, r.median, r.name, r.quantile,
#r.kurt, r.cov, r.corr, r.aggregate, 
# r.std, r.skew, r.sum, r.var
df.plot( figsize=(10,4))
r.mean().plot(figsize=(10,4))
#exponentially weighted(指数加权)
df.ewm(min_periods=2, adjust=True, span=50, ignore_na=False).mean().plot(figsize=(10,4))

# 累计求和
df = df.cumsum()

df.rolling(window = 50).sum().plot(figsize=(10,4), title='cumsum')

# 自定义函数
df.rolling(window =50).apply(lambda x: np.fabs(x - x.mean()).mean()).plot(figsize=(10,4))

df.head()

	A	B	C
2016-07-01	1.251741	-0.669649	0.845530
2016-07-02	0.186175	-0.269975	0.810724
2016-07-03	-0.804764	-0.484257	-1.079426
2016-07-04	1.280105	0.557200	-0.154886
2016-07-05	3.042824	-0.878534	-1.389216

Expanding windows

截止每个时间点累计求和的平均值

df.expanding(min_periods = 1).mean().plot(figsize=(10,4))