Python Pandas if语句基于group by sum

Question

使用这个python pandas dataframe df：

CategoryA | CategoryB | Count
1           A           0
1           A           -1
2           B           1
2           B           1
3           C           1
3           C           -1

我基本上想要删除所有ClassA / B的分组，其总和低于0.

df['decision'] = np.where(df.groupby(['CategoryA', 'CategoryB'])['Count'].sum()>0, 'keep', 'delete')

但我收到此错误 ValueError：值的长度与索引的长度不匹配

输出将是：

CategoryA | CategoryB | Count | decision
1           A           0       delete
1           A           -1      delete
2           B           1       keep
2           B           1       keep
3           C           1       delete
3           C           -1      delete

更愿意使用df.loc执行此操作，但不确定如何使用。

Answer 1

In [67]: df['decision'] = \
             np.where(df.groupby(['CategoryA', 'CategoryB'])['Count'].transform('sum') > 0, 
                      'keep', 'delete')

In [68]: df
Out[68]:
   CategoryA CategoryB  Count decision
0          1         A      0   delete
1          1         A     -1   delete
2          2         B      1     keep
3          2         B      1     keep
4          3         C      1   delete
5          3         C     -1   delete

Answer 2

你走在正确的轨道上。

m = df.groupby(['CategoryA', 'CategoryB']).transform('sum').gt(0)
df['decision'] = np.where(m, 'keep', 'delete')

df
   CategoryA CategoryB  Count decision
0          1         A      0   delete
1          1         A     -1   delete
2          2         B      1     keep
3          2         B      1     keep
4          3         C      1   delete
5          3         C     -1   delete

使用transform检索大小相同的结果。

Answer 3

df['decision']=df['CategoryB'].map(df.groupby('CategoryB')['Count'].\
      apply(lambda x :np.where(x.sum()>0,'keep','delete')))
df
Out[573]: 
   CategoryA CategoryB  Count decision
0          1         A      0   delete
1          1         A     -1   delete
2          2         B      1     keep
3          2         B      1     keep
4          3         C      1   delete
5          3         C     -1   delete