python pandas重采样计数和总和

Question

我有按日期计算的数据，并希望按周创建一个新的数据框，其中包含销售额总和和类别计数。

#standard packages
import numpy as np
import pandas as pd

#visualization
%matplotlib inline
import matplotlib.pylab as plt

#create weekly datetime index
edf = pd.read_csv('C:\Users\j~\raw.csv', parse_dates=[6])
edf2 = edf[['DATESENT','Sales','Category']].copy()
edf2

#output

DATESENT    |  SALES  | CATEGORY
2014-01-04      100        A
2014-01-05      150        B
2014-01-07      150        C
2014-01-10      175        D

#create datetime index of week
edf2['DATESENT']=pd.to_datetime(edf2['DATESENT'],format='%m/%d/%Y')
edf2 = edf2.set_index(pd.DatetimeIndex(edf2['DATESENT']))
edf2.resample('w').sum()
edf2

#output

            SALES CATEGORY 
DATESENT     
2014-01-05  250      AB
2014-01-12  325      CD

但我正在寻找

           SALES CATEGORY 
DATESENT     
2014-01-05  250      2
2014-01-12  325      2

这没用...

edf2 = e2.resample('W').agg("Category":len,"Sales":np.sum)

谢谢

Answer 1

Agg 将字典作为 various formats 中的参数。

edf2 = e2.resample('W').agg({"Category":'size',"Sales":'sum'})

Answer 2

使用pd.TimeGrouper + agg

f = {'SALES': 'sum', 'CATEGORY': 'count'}
g = pd.TimeGrouper('W')
df.set_index('DATESENT').groupby(g).agg(f)

            CATEGORY  SALES
DATESENT                   
2014-01-05         2    250
2014-01-12         2    325