如何在给定的不规则日期重新采样时间序列

Question

import pandas as pd
date_index = pd.date_range("2010-01-31", "2010-12-31", freq="M")
df  = pd.Series(range(12), index=date_index)

dates = date_index[1::2]

The Series df is of monthly frequency, and we want to resample by adding up the value between the dates as given by the dates variable.

df is:

2010-01-31     0
2010-02-28     1
2010-03-31     2
2010-04-30     3
2010-05-31     4
2010-06-30     5
2010-07-31     6
2010-08-31     7
2010-09-30     8
2010-10-31     9
2010-11-30    10
2010-12-31    11
Freq: M, dtype: int64

dates is

DatetimeIndex(['2010-02-28', '2010-04-30', '2010-06-30', '2010-08-31',
               '2010-10-31', '2010-12-31'],
              dtype='datetime64[ns]', freq='2M')

The expected result should be:

2010-02-28     1
2010-04-30     5
2010-06-30     9
2010-08-31     13
2010-10-31     17
2010-12-31    21

Answer 1

Not a general resampling solution but for your concrete question of adding up the values between the dates you could use

res = df.cumsum()[dates].diff()
res[0] = df[dates[0]]
res = res.astype(df.dtype)

Result:

2010-02-28     1
2010-04-30     5
2010-06-30     9
2010-08-31    13
2010-10-31    17
2010-12-31    21

Answer 2

Idea is replace not matched values of date s to missing values by Series.where with bfill for bacj filling missing values and then aggregate sum :

date_index = pd.date_range("2010-01-31", "2010-12-31", freq="M")
s  = pd.Series(range(12), index=date_index)

dates = date_index[1::2]

a = s.index.to_series().where(s.index.isin(dates)).bfill()
out = s.groupby(a).sum()
print(out)
2010-02-28     1
2010-04-30     5
2010-06-30     9
2010-08-31    13
2010-10-31    17
2010-12-31    21
dtype: int64

Answer 3

For your specific example, where df[0] = 0 , it is a simple resample with sum() aggregation, skipping df[0] .

df_resampled = df[1::].resample('2M').sum()

print(df_resampled)
2010-02-28     1
2010-04-30     5
2010-06-30     9
2010-08-31    13
2010-10-31    17
2010-12-31    21
Freq: 2M, dtype: int64

In case df[0] != 0 , you can still make an easy workaround by adding df[0] to the first element of df_resampled :

df_resampled[0] = df_resampled[0] + df[0]

In case you want general resampling with period of two month, you can try to use the loffset parameter of resample and provide a function returning pd.Timedelta objects such, that it "floors" to the last day of each individual month. (See here for how to get montly periods for pd.Timedelta )