将多级索引的一个级别拆分为多个列

问题描述

所以我有一个数据框:

df = pd.DataFrame([["foo","fizz",1],["foo","fizz",2],["foo","buzz",3],["foo","buzz",4],["bar","fizz",6],["bar","buzz",8]],columns=["a","b","c"])

       a    b     c
    0  foo  fizz  1
    1  foo  fizz  2
    2  foo  buzz  3
    3  foo  buzz  4
    4  bar  fizz  6
    5  bar  buzz  8

我可以分组:

df2 = df.groupby(["a","b"]).sum()

              c
    a   b
    bar buzz  8
        fizz  6
    foo buzz  7
        fizz  3

哪个很棒！但是我真正需要的是两列，而不是"c"列:"foo"和"bar":

Which is awesome! But what I really need, instead of the "c" column is two columns, "foo" and "bar":

          foo  bar
    b
    buzz  7    8
    fizz  3    6

有人可以建议一种方法吗?我尝试搜索，但是我想我没有正确的术语，所以我什么也找不到.

Can someone suggest a way to do this? I tried searching, but I guess I don't have the correct terminology for this so I couldn't find anything.

推荐答案

您可以为此使用unstack:

df2.unstack(level='a')

示例:

In [146]: df2.unstack(level='a')
Out[146]:
       c
a    bar foo
b
buzz   8   7
fizz   6   3

之后，您将获得多索引列.如果需要获取平面数据框，则可以使用multiindex的droplevel:

After that you'll get multiindexed columns. If you need to get flat dataframe you could use droplevel of multiindex:

df3 = df2.unstack(level='a')
df3.columns = df3.columns.droplevel()

In [177]: df3
Out[177]:
a     bar  foo
b
buzz    8    7
fizz    6    3

编辑

droplevel从MultiIndex降低级别，该列在unstack之后变为.默认情况下，它删除级别0，这是该数据帧所需的级别.

droplevel drops level from MultiIndex which your columns become after unstack. By default it drops level 0 which is what you need for that dataframe.

从help(pd.core.index.MultiIndex.droplevel)复制:

下降级别(自身，级别= 0) 返回索引，删除了请求的级别.如果MultiIndex只有2 级别，结果将是索引类型而不是MultiIndex.

droplevel(self, level=0) Return Index with requested level removed. If MultiIndex has only 2 levels, the result will be of Index type not MultiIndex.

Parameters
----------
level : int/level name or list thereof

Notes
-----
Does not check if result index is unique or not

Returns
-------
index : Index or MultiIndex

这篇关于将多级索引的一个级别拆分为多个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！