python - 在 Pandas 中，如何仅使用 True 系列 bool 值对行进行求和

Question

这个问题是 moys 问题的延伸，因为我对如何根据 bool 值的真值序列进行求和的答案感兴趣。假设我有这个数据框，我只想对真实行进行求和。:

   id log   loc  pos_evnts  neg_evnts   As  non_As  pos_wrds  neg_wrds  As/Ac  Truth  T
0   A   c  City          8          0   48       0         0         0      1  False  1
1   A   d  City          2          6    0     180         4        10      0   True  2
2   A   e  City          0         22   87       0         0         0      1   True  2
3   A   f  City          8          0   35       0         0         0      1  False  3
4   A   g  City          8          2   42       0         0         0      1  False  3
5   A   h  City          4          4    0     115         4         2      0   True  4
6   A   i  City          2          0   32       0         0         0      1   True  4
7   B   j  Hill          3          0   24       0         0         0      1  False  5
8   B   k  City          6          8  116       0         0         2      1  False  5
9   B   l  City          2          4  200       0         0         2      1  False  5
10  C   m  City          2          0   40       0         0         0      0   True  6
11  C   n  Hill          5          0    1       0         2         0      0   True  6
12  C   o  City          5          0    7       0         0         5      1   True  6

我想对行进行求和以获得这个答案(真实的行是求和的):

    pos_evnts  neg_evnts   As  non_As  pos_wrds  neg_wrds  As/Ac  
0           8          0   48       0         0         0      1 
1           2          6    0     180         4        10      0  
2           2         28   87     180         4        10      1  
3           8          0   35       0         0         0      1  
4           8          2   42       0         0         0      1 
5           4          4    0     115         4         2      0   
6           6          4   32     115         4         2      1   
7           3          0   24       0         0         0      1  
8           6          8  116       0         0         2      1 
9           2          4  200       0         0         2      1  
10          2          0   40       0         0         0      0  
11          7          0   41       0         2         0      0  
12         12          0   48       0         2         5      1  

我尝试过:

df.groupby((df['T'])).cumsum()

In [4738]: df.groupby(df['T']).cumsum()                                                                                                                                                        
Out[4738]: 
    pos_evnts  neg_evnts   As  non_As  pos_wrds  neg_wrds  As/Ac  Truth
0           8          0   48       0         0         0      1  0.000
1           2          6    0     180         4        10      0  1.000
2           2         28   87     180         4        10      1  2.000
3           8          0   35       0         0         0      1  0.000
4          16          2   77       0         0         0      2  0.000
5           4          4    0     115         4         2      0  1.000
6           6          4   32     115         4         2      1  2.000
7           3          0   24       0         0         0      1  0.000
8           9          8  140       0         0         2      2  0.000
9          11         12  340       0         0         4      3  0.000
10          2          0   40       0         0         0      0  1.000
11          7          0   41       0         2         0      0  2.000
12         12          0   48       0         2         5      1  3.000

但它是错误的(真相:0.000 行)。我希望它只对真实行进行求和。任何帮助，将不胜感激。如何修改公式以忽略 cumsum 的 False 行。

Answer 1

您可以仅过滤包含数字列的 True 行，也排除 T 列以防止累积总和并分配回来:

cols = df.select_dtypes(np.number).columns.difference(['T'])
df.loc[df['Truth'], cols] = df.loc[df['Truth'], cols] .groupby(df['T']).cumsum() 
print (df)
   id log   loc  pos_evnts  neg_evnts   As  non_As  pos_wrds  neg_wrds  As/Ac  \
0   A   c  City          8          0   48       0         0         0      1   
1   A   d  City          2          6    0     180         4        10      0   
2   A   e  City          2         28   87     180         4        10      1   
3   A   f  City          8          0   35       0         0         0      1   
4   A   g  City          8          2   42       0         0         0      1   
5   A   h  City          4          4    0     115         4         2      0   
6   A   i  City          6          4   32     115         4         2      1   
7   B   j  Hill          3          0   24       0         0         0      1   
8   B   k  City          6          8  116       0         0         2      1   
9   B   l  City          2          4  200       0         0         2      1   
10  C   m  City          2          0   40       0         0         0      0   
11  C   n  Hill          7          0   41       0         2         0      0   
12  C   o  City         12          0   48       0         2         5      1   

    Truth  T  
0   False  1  
1    True  2  
2    True  2  
3   False  3  
4   False  3  
5    True  4  
6    True  4  
7   False  5  
8   False  5  
9   False  5  
10   True  6  
11   True  6  
12   True  6