pandas 聚合-如何保留所有列 [英] Pandas aggregate -- how to retain all columns

问题描述

示例数据帧：

rand = np.random.RandomState(1)
df = pd.DataFrame({'A': ['group1', 'group2', 'group3'] * 2,
                'B': rand.rand(6),
                'C': rand.rand(6),
                'D': rand.rand(6)})

print df

        A         B         C         D
0  group1  0.417022  0.186260  0.204452
1  group2  0.720324  0.345561  0.878117
2  group3  0.000114  0.396767  0.027388
3  group1  0.302333  0.538817  0.670468
4  group2  0.146756  0.419195  0.417305
5  group3  0.092339  0.685220  0.558690

Groupby列A

Groupby column A

group = df.groupby('A')

使用agg返回每个组的最大值

Use agg to return max value for each group

max1 = group['B'].agg({'max' : np.max})
print max1

             max
A               
group1  0.417022
group2  0.720324
group3  0.092339

但我想保留（或取回）适当的数据在其他列C和D中。这将是包含最大值的行的剩余数据。
因此，返回值应为：

But I would like to retain (or get back) the appropriate data in the other columns, C and D. This would be the remaining data for the row which contained the max value. So, the return should be:

     A         B         C         D
group1  0.417022  0.186260  0.204452
group2  0.720324  0.345561  0.878117
group3  0.092339  0.685220  0.558690

有人可以显示如何执行此操作吗？

Can anybody show how to do this? Any help appreciated.

推荐答案

两个阶段：首先查找索引，然后查找所有行。

Two stages: first find indices, then lookup all the rows.

idx = df.groupby('A').apply(lambda x: x['B'].argmax())
idx

Out[362]: 
A
group1    0
group2    1
group3    5

df.loc[idx]

Out[364]: 
        A         B         C         D
0  group1  0.417022  0.186260  0.204452
1  group2  0.720324  0.345561  0.878117
5  group3  0.092339  0.685220  0.558690

这篇关于 pandas 聚合-如何保留所有列的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！