在pandas数据框中的groupby之后选择每个组中的前3个类别

Question

so my dataframe looks like this now:

| Name | Type | Class   | Amount |
|------|------|---------|--------|
| Abel | A    | Chinese | 2      |
| Abel | B    | English | 5      |
| Abel | C    | Science | -1     |
| Abel | D    | Physics | -10    |
| Cain | C    | Chinese | -5     |
| Cain | B    | Science | 0      |
| Cain | A    | English | 30     |
| Cain | D    | Chinese | 2      |
|------|------|---------|--------|

data sample:

data = {'Name': ['Abel', 'Abel', 'Abel', 'Abel', 'Cain', 'Cain', 'Cain', 'Cain'],
'Type': ['A', 'B', 'C', 'D', 'C', 'B', 'A', 'D'],
'Class': ['Chinese', 'English', 'Science', 'Physics', 'Chinese', 'Science', 'English', 'Chinese'],
'Amount': [2,5,-1,-10,-5,0,30,2]}

I'm trying to find for each name what are the top n type and top n class based on the amount.

I tried df.groupby(["Name","Type"]).sum() which gives me the groupings but how can I select the top 5 so that I can transpose them into 5 different columns?

Eg The final output for top 3 types should be something like this, top 3 classes is just something similar except the columns are class 1 to class 3:

| Name | Type 1 | Type 2 | Type 3 |
|------|--------|--------|--------|
| Abel | B      |   A    |   C    |
| Cain | A      |   D    |   B    |

I tried sort_values and then .head(5) too but somehow the sort treats negative numbers as the biggest. Besides, the results also fall out of grouping. Any help please? Thanks!

Answer 1

Use:

#sorting by both columns
df1 = df.sort_values(['Name','Amount'], ascending=[True, False])
#create counter column used for later columns names
df1['g'] = df1.groupby('Name').cumcount().add(1)
#filter top3
df1 = df1[df1['g'] <= 3]
#reshape by pivot
df2 = df1.pivot('Name','g','Type').add_prefix('Type ').reset_index().rename_axis(None, axis=1)
print (df2)
   Name Type 1 Type 2 Type 3
0  Abel      B      A      C
1  Cain      A      D      B

Answer 2

I'm not 100% sure if I understand your question correctly, but you can use

df.nlargest(5,["Amount"])

This will select the 5 largest amounts. You can adjust the 5.