带有 pct_change 的 Pandas groupby
声明：本页面是StackOverFlow热门问题的中英对照翻译，遵循CC BY-SA 4.0协议，如果您需要使用它，必须同样遵循CC BY-SA许可，注明原文地址和作者信息，同时你必须将它归于原作者（不是我）：StackOverFlow
原文地址: http://stackoverflow.com/questions/40273251/
Warning: these are provided under cc-by-sa 4.0 license. You are free to use/share it, But you must attribute it to the original authors (not me):
StackOverFlow
Pandas groupby with pct_change
提问by user3357979
I'm trying to find the period over period growth in value for each unique group, grouped by Company, Group, and Date.
我试图找到按公司、组和日期分组的每个独特组的期间价值增长。
Company Group Date Value
A X 2015-01 1
A X 2015-02 2
A X 2015-03 1.5
A XX 2015-01 1
A XX 2015-02 1.5
A XX 2015-03 0.75
A XX 2015-04 1
B Y 2015-01 1
B Y 2015-02 1.5
B Y 2015-03 2
B Y 2015-04 3
B YY 2015-01 2
B YY 2015-02 2.5
B YY 2015-03 3
I've tried:
我试过了：
df.groupby(['Date','Company','Group']).pct_change()
but this returns all NaN.
但这会返回所有 NaN。
The result I'm looking for is:
我正在寻找的结果是：
Company Group Date Value/People
A X 2015-01 NaN
A X 2015-02 1.0
A X 2015-03 -0.25
A XX 2015-01 NaN
A XX 2015-02 0.5
A XX 2015-03 -0.5
A XX 2015-04 0.33
B Y 2015-01 NaN
B Y 2015-02 0.5
B Y 2015-03 0.33
B Y 2015-04 0.5
B YY 2015-01 NaN
B YY 2015-02 0.25
B YY 2015-03 0.2
回答by piRSquared
you want to get your date into the row index and groups/company into the columns
您想将日期放入行索引并将组/公司放入列中
d1 = df.set_index(['Date', 'Company', 'Group']).Value.unstack(['Company', 'Group'])
d1
then use pct_change
然后使用 pct_change
d1.pct_change()
OR
或者
with groupby
与 groupby
df['pct'] = df.sort_values('Date').groupby(['Company', 'Group']).Value.pct_change()
df
回答by SimonR
I'm not sure the groupbymethod works as intended as of Pandas 0.23.4 at least.
我不确定该groupby方法至少在 Pandas 0.23.4 中是否按预期工作。
df['pct'] = df.sort_values('Date').groupby(['Company', 'Group']).Value.pct_change()
Produces this, which is incorrect for purposes of the question:
产生这个，这对于问题的目的是不正确的：
The Index+Stack method still works as intended, but you need to do additional merges to get it into the original form requested.
Index+Stack 方法仍然按预期工作，但您需要进行额外的合并以使其成为请求的原始形式。
d1 = df.set_index(['Date', 'Company', 'Group']).Value.unstack(['Company', 'Group'])
d1 = d1.pct_change().stack([0,1]).reset_index()
df = df.merge(d1, on=['Company', 'Group', 'Date'], how='left')
df.rename(columns={0: 'pct'}, inplace=True)
df
