关于这个问题 Grouping Pandas dataframe across rows ，操作是:
amount
clients
Comp1 16.360417
Comp2 69.697501
Comp3 85.700000
Comp4 36.666667
Comp5 44.156500
如果将日期列添加到输入中:
tdate,client1,client2,client3,client4,client5,client6,amount
12/31/2017,,,Comp1,,,4.475000
12/31/2017,,,Comp2,,,16.305584
10/31/2107,,,Comp3,,,4.050000
10/31/2017,Comp2,Comp1,,Comp4,,,21.000000
1/1/2017,,,Comp4,,,30.000000
2/2/2017,Comp1,,Comp2,,,5.137500
10/31/2017,,,Comp3,,,52.650000
12/31/2017,,,Comp1,,,2.650000
10/31/2017,Comp3,,,Comp3,,,29.000000
12/31/2017,Comp5,,,Comp2,,,20.809000
1/1/2017,Comp5,,,Comp2,,,15.100000
10/31/2017,Comp5,,,Comp2,,,52.404000
我们如何得到这个输出:
12/31/2017 Comp1 4.475+2.65
12/31/2017 Comp2 16.305584+20.809/2
10/31/2017 Comp2 21/3+5.1375/2+52.404/2
1/1/2017 Comp2 15.1/2
10/31/2017 Comp3 4.05+52.65+29
1/1/2017 Comp4 30
10/21/2017 Comp4 21/3
12/31/2017 Comp5 20.809/2
1/1/2017 Comp5 15.1/2
10/31/2017 Comp5 52.404/2
最佳答案
改进之前的答案，我们需要通过设置两列作为索引来使用堆栈。
cols= ['amount','tdate']
df['new'] = df['amount']/df.drop(cols,1).count(1)
#Set the index as new and tdate by droping amount column, stack and drop the nans.
x = df.drop(['amount'],1).set_index(['new','tdate']).stack().dropna()
#Create dataframe from amount,tdate and the clients
ndf = pd.DataFrame({'amount':x.index.get_level_values('new'),'tdate':x.index.get_level_values('tdate'),'clients':x.values})
#Groupby `clients` and `tdate`
ndf.groupby(['clients','tdate']).sum().reset_index()
输出:
clients tdate amount 0 Comp1 10/31/2017 7.000000 1 Comp1 12/31/2017 7.125000 2 Comp1 2/2/2017 2.568750 3 Comp2 1/1/2017 7.550000 4 Comp2 10/31/2017 33.202000 5 Comp2 12/31/2017 26.710084 6 Comp2 2/2/2017 2.568750 7 Comp3 10/31/2017 81.650000 8 Comp3 10/31/2107 4.050000 9 Comp4 1/1/2017 30.000000 10 Comp4 10/31/2017 7.000000 11 Comp5 1/1/2017 7.550000 12 Comp5 10/31/2017 26.202000 13 Comp5 12/31/2017 10.404500
关于python - 跨行对 Pandas 数据框进行分组 - 2.0，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/46997707/