您需要DataFrameGroupBy.idxmax
选择的最大值为value3
和的索引DataFrame
由loc
:
print (df.groupby(['id1','id2','value1']).value3.idxmax())
id1 id2 value1
1 2 30 1
3 5 12 4
24 12 1 6
Name: value3, dtype: int64
df = df.loc[df.groupby(['id1','id2','value1']).value3.idxmax()]
print (df)
id1 id2 value1 value2 value3 a
1 1 2 30 42 26.2 NaN
4 3 5 12 33 11.2 NaN
6 24 12 1 23 1.9 NaN
另一种可能的解决方案是sort_values
按列value3
,然后groupby
使用GroupBy.first
:
df = df.sort_values('value3', ascending=False)
.groupby(['id1','id2','value1'], sort=False)
.first()
.reset_index()
print (df)
id1 id2 value1 value2 value3 a
0 1 2 30 42 26.2 NaN
1 3 5 12 33 11.2 NaN
2 24 12 1 23 1.9 NaN