问题
I have a question regarding the resampling method of pandas Dataframes. I have a DataFrame with one observation per day:
import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randint(0,100,size=(366, 1)), columns=list('A'))
df.index = pd.date_range(datetime.date(2016,1,1),datetime.date(2016,12,31))
if I want to compute the sum (or other) for every month, I can directly do:
EOM_sum = df.resample(rule="M").sum()
however I have a specific calendar (irregular frequency):
import datetime
custom_dates = pd.DatetimeIndex([datetime.date(2016,1,13),
datetime.date(2016,2,8),
datetime.date(2016,3,16),
datetime.date(2016,4,10),
datetime.date(2016,5,13),
datetime.date(2016,6,17),
datetime.date(2016,7,12),
datetime.date(2016,8,11),
datetime.date(2016,9,10),
datetime.date(2016,10,9),
datetime.date(2016,11,14),
datetime.date(2016,12,19),
datetime.date(2016,12,31)])
If I want to compute the sum for each period, I currently add a temporary column to df with the end of the period each row belongs to, and then perform the operation with a groupby:
df["period"] = custom_dates[custom_dates.searchsorted(df.index)]
custom_sum = df.groupby(by=['period']).sum()
However this is quite dirty and doesn't look pythonic. Is there a built-in method to do this in Pandas? Thanks in advance.
回答1:
Creating nw column is not necessary, you can groupby by DatatimeIndex, because length is same as lenght of df:
import pandas as pd
import numpy as np
np.random.seed(100)
df = pd.DataFrame(np.random.randint(0,100,size=(366, 1)), columns=list('A'))
df.index = pd.date_range(datetime.date(2016,1,1),datetime.date(2016,12,31))
print (df.head())
A
2016-01-01 8
2016-01-02 24
2016-01-03 67
2016-01-04 87
2016-01-05 79
import datetime
custom_dates = pd.DatetimeIndex([datetime.date(2016,1,13),
datetime.date(2016,2,8),
datetime.date(2016,3,16),
datetime.date(2016,4,10),
datetime.date(2016,5,13),
datetime.date(2016,6,17),
datetime.date(2016,7,12),
datetime.date(2016,8,11),
datetime.date(2016,9,10),
datetime.date(2016,10,9),
datetime.date(2016,11,14),
datetime.date(2016,12,19),
datetime.date(2016,12,31)])
custom_sum = df.groupby(custom_dates[custom_dates.searchsorted(df.index)]).sum()
print (custom_sum)
A
2016-01-13 784
2016-02-08 1020
2016-03-16 1893
2016-04-10 1242
2016-05-13 1491
2016-06-17 1851
2016-07-12 1319
2016-08-11 1348
2016-09-10 1616
2016-10-09 1523
2016-11-14 1793
2016-12-19 1547
2016-12-31 664
Another solution is append new index by custom_dates, groupby use numpy array as output from searchsorted function:
print (custom_dates.searchsorted(df.index))
[ 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4
4 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5
5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6
6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 7 7 7 7 7 7
7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 8
8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8
8 8 8 8 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9
9 9 9 9 9 9 9 9 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 11 11 11 11 11 11
11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11
11 11 11 11 12 12 12 12 12 12 12 12 12 12 12 12]
custom_sum = df.groupby(custom_dates.searchsorted(df.index)).sum()
custom_sum.index = custom_dates
print (custom_sum)
A
2016-01-13 784
2016-02-08 1020
2016-03-16 1893
2016-04-10 1242
2016-05-13 1491
2016-06-17 1851
2016-07-12 1319
2016-08-11 1348
2016-09-10 1616
2016-10-09 1523
2016-11-14 1793
2016-12-19 1547
2016-12-31 664
来源：https://stackoverflow.com/questions/41854000/pandas-dataframe-resampling-with-specific-dates