[英]Pandas pivot table with multiple columns at once
I wonder if pandas.pivot_table
can accept two columns at once and process them separately instead of hierarchically. 我想知道pandas.pivot_table
可以一次接受两列并分别处理而不是按层次处理。
Say I have the following data frame: 说我有以下数据框:
id date day val
101 11/1/1 1 2.1
101 11/1/2 2 2.2
101 11/1/3 3 2.3
102 11/1/2 1 3.1
102 11/1/3 2 3.2
102 11/1/4 3 3.3
I want the result to be like this: 我希望结果是这样的:
date day
id 11/1/1 11/1/2 11/1/3 11/1/4 1 2 3
101 2.1 2.2 2.3 NaN 2.1 2.2 2.3
102 NaN 3.1 3.2 3.3 3.1 3.2 3.3
When I do df.pivot_table(index='id', columns=['date','day'], values='val')
, it will integrate date
and day
into a hierarchy which is not what I want. 当我执行df.pivot_table(index='id', columns=['date','day'], values='val')
,它将把date
和day
集成到我不想要的层次结构中。 Of course I can do twice with date
and day
respectively and concatenate the results, but is there a more convenient way to do so at once? 当然,我可以做两次, date
和day
分别连接结果,但有一个更便捷的方式这样做一次?
You can make 2 pivot
calls and concat
enate the result. 你可以让2个pivot
电话和concat
enate结果。
i = df.pivot('id', 'date', 'val')
j = df.pivot('id', 'day', 'val')
pd.concat([i, j], 1, keys=['date', 'day'])
date day
11/1/1 11/1/2 11/1/3 11/1/4 1 2 3
id
101 2.1 2.2 2.3 NaN 2.1 2.2 2.3
102 NaN 3.1 3.2 3.3 3.1 3.2 3.3
As a single liner - 作为单个衬板-
c = ['date', 'day'] # add more cols as needed
pd.concat([df.pivot('id', x, 'val') for x in c], axis=1, keys=c)
date day
11/1/1 11/1/2 11/1/3 11/1/4 1 2 3
id
101 2.1 2.2 2.3 NaN 2.1 2.2 2.3
102 NaN 3.1 3.2 3.3 3.1 3.2 3.3
One line ....but why you need them in one line.. 一行....但是为什么需要一行
df.set_index(['id','val']).stack().to_frame('col').\
reset_index(level='val').set_index('col',append=True).\
unstack([-2,-1]).sort_index(1,level=1)
Out[69]:
val
date day
col 11/1/1 11/1/2 11/1/3 11/1/4 1 2 3
id
101 2.1 2.2 2.3 NaN 2.1 2.2 2.3
102 NaN 3.1 3.2 3.3 3.1 3.2 3.3
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.