Python 按相似的非匹配值对数据帧进行分组

逻辑并不完全清楚，但假设你想要在差距超过10%的情况下开始一个新的小组:

close = (df.sort_values(by=['amount', 'num'])
           .groupby('amount')
           ['num'].diff().abs().gt(10).cumsum()
         )

for _, g in df.groupby(['amount', close]):
    print(g, end='\n\n')

输出:

        id  amount  num
0  aaa-aaa     130   12
3  ddd-ddd     130   19

        id  amount  num
1  bbb-bbb     130   39

        id  amount  num
2  ccc-ccc     270   41
4  eee-eee     270   37

how it works:

# sort values by amount/sum
df.sort_values(by=['amount', 'num'])

        id  amount  num
0  aaa-aaa     130   12
3  ccc-ccc     130   19
1  aaa-aaa     130   39
5  aaa-aaa     130   42
4  bbb-bbb     270   37
2  bbb-bbb     270   41
6  aaa-aaa     380   39

# get the absolute successive difference in "num"
(df.sort_values(by=['amount', 'num'])
   .groupby('amount')
   ['num'].diff().abs()
)

0     NaN
3     7.0
1    20.0
5     3.0
4     NaN
2     4.0
6     NaN
Name: num, dtype: float64

# check if it's greater than 10 and cumsum
# to create a grouper for groupby

[...].gt(10).cumsum()

0    0
3    0
1    1
5    1
4    1
2    1
6    1
Name: num, dtype: int64