python - Groupby 对象按组的百分比变化

Question

我需要 pandas groupby 方面的帮助。有没有办法对 pandas groupby 中的每个组运行 lambda(或等效函数)？请参见下面的示例。我想将上一年的百分比变化添加到此 groupby 右侧的列中。我尝试了几种方法，但它们似乎都忽略了从新的“项目”组重新开始。

import pandas as pd
x = pd.Series(['Oranges', 'Apples', 'Other Fruits', 'Oranges', 'Apples', 'Other Fruits', 'Oranges', 'Apples', 'Other Fruits'])
y = pd.Series([2016, 2016, 2016, 2017, 2017, 2017, 2018, 2018, 2018])
z = pd.Series([12, 15, 9, 14, 15, 50, 32, 15, 12])
df = pd.DataFrame({'Item': x, 'Year':y, 'Values':z})
df=df.sort_values('Values', ascending=False) 
df.groupby(['Item', 'Year']).sum()
#How do I get Percent % Values for each group as a new column right of 'Values'

我期待以下内容:

Answer 1

你正在寻找 GroupBy + apply with pct_change:

# Sort DataFrame before grouping.
df = df.sort_values(['Item', 'Year']).reset_index(drop=True)
# Group on keys and call `pct_change` inside `apply`.
df['Change'] = df.groupby('Item', sort=False)['Values'].apply(
     lambda x: x.pct_change()).to_numpy()

df
           Item  Year  Values    Change
0        Apples  2016      15       NaN
1        Apples  2017      15  0.000000
2        Apples  2018      15  0.000000
3       Oranges  2016      12       NaN
4       Oranges  2017      14  0.166667
5       Oranges  2018      32  1.285714
6  Other Fruits  2016       9       NaN
7  Other Fruits  2017      50  4.555556
8  Other Fruits  2018      12 -0.760000