根据另一列 pandas 查找前 5 个值

Question

How to find top 5 values of category column based while grouping customer_id column in pandas dataframe?

       customer_id     email                        address_id  name              category

0         411   NORMAN.CURRIER@sakilacustomer.org       416     NORMAN CURRIER      Scifi
1         411   NORMAN.CURRIER@sakilacustomer.org       416     NORMAN CURRIER      Action
2         411   NORMAN.CURRIER@sakilacustomer.org       416     NORMAN CURRIER      Sports
3         411   NORMAN.CURRIER@sakilacustomer.org       416     NORMAN CURRIER      Scifi
4         411   NORMAN.CURRIER@sakilacustomer.org       416     NORMAN CURRIER      Family
5         411   NORMAN.CURRIER@sakilacustomer.org       416     NORMAN CURRIER      Action
6         527   CORY.MEEHAN@sakilacustomer.org          533     CORY MEEHAN         Documentary
7         527   CORY.MEEHAN@sakilacustomer.org          533     CORY MEEHAN         Action
8         527   CORY.MEEHAN@sakilacustomer.org          533     CORY MEEHAN         Sports
9         527   CORY.MEEHAN@sakilacustomer.org          533     CORY MEEHAN         Scifi
10        527   CORY.MEEHAN@sakilacustomer.org          533     CORY MEEHAN         Documentary
11        527   CORY.MEEHAN@sakilacustomer.org          533     CORY MEEHAN         Sports

I want another column named preferred_film_category for each unique customer_id ( the top 5 values are based on how many times particular category occurs for each unique customer_id )

Expected Dataframe:

       customer_id     email                     address_id    name       category      preferred_film_category   

0       411    NORMAN.CURRIER@sakilacustomer.org   416   NORMAN CURRIER    Scifi        Scifi, Action, Sports, Animation, Drama    
1       411    NORMAN.CURRIER@sakilacustomer.org   416   NORMAN CURRIER    Action       Scifi, Action, Sports, Animation, Drama 
2       411    NORMAN.CURRIER@sakilacustomer.org   416   NORMAN CURRIER    Sports       Scifi, Action, Sports, Animation, Drama 
3       411    NORMAN.CURRIER@sakilacustomer.org   416   NORMAN CURRIER    Scifi        Scifi, Action, Sports, Animation, Drama 
4       411    NORMAN.CURRIER@sakilacustomer.org   416   NORMAN CURRIER    Family       Scifi, Action, Sports, Animation, Drama 
5       411    NORMAN.CURRIER@sakilacustomer.org   416   NORMAN CURRIER    Action       Scifi, Action, Sports, Animation, Drama
6       527     CORY.MEEHAN@sakilacustomer.org     533   CORY MEEHAN       Documentary  Documentary, Sports, Scifi, Action 
7       527     CORY.MEEHAN@sakilacustomer.org     533   CORY MEEHAN       Action       Documentary, Sports, Scifi, Action 
8       527     CORY.MEEHAN@sakilacustomer.org     533   CORY MEEHAN       Sports       Documentary, Sports, Scifi, Action 
9       527     CORY.MEEHAN@sakilacustomer.org     533   CORY MEEHAN       Scifi        Documentary, Sports, Scifi, Action 
10      527     CORY.MEEHAN@sakilacustomer.org     533   CORY MEEHAN       Documentary  Documentary, Sports, Scifi, Action 
11      527     CORY.MEEHAN@sakilacustomer.org     533   CORY MEEHAN       Sports       Documentary, Sports, Scifi, Action 

Answer 1

Try value_counts + groupby nlargest to get get the highest frequency categories, then groupby aggregate to convert to a string, then join to merge back with the original DataFrame:

n = 2
df = df.join(
    df.value_counts(['customer_id', 'category'])
        .groupby(level=0).nlargest(n)
        .reset_index('category')
        .groupby(level=0)['category'].agg(', '.join)
        .rename('preferred_film_category'),
    on='customer_id'
)

df :

    customer_id                              email  address_id            name     category preferred_film_category
0           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER        Scifi           Action, Scifi
1           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Action           Action, Scifi
2           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Sports           Action, Scifi
3           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER        Scifi           Action, Scifi
4           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Family           Action, Scifi
5           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Action           Action, Scifi
6           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN  Documentary     Documentary, Sports
7           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN       Action     Documentary, Sports
8           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN       Sports     Documentary, Sports
9           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN        Scifi     Documentary, Sports
10          527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN  Documentary     Documentary, Sports
11          527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN       Sports     Documentary, Sports

*note n is set to 2 as each customer only has 4 unique values in categrory and so 5 does not demonstrate the functioning of the code. Change this to the desired value to keep ( 5 ).

---

DataFrame Used:

df = pd.DataFrame({
    'customer_id': [411, 411, 411, 411, 411, 411, 527, 527, 527, 527, 527, 527],
    'email': ['NORMAN.CURRIER@sakilacustomer.org',
              'NORMAN.CURRIER@sakilacustomer.org',
              'NORMAN.CURRIER@sakilacustomer.org',
              'NORMAN.CURRIER@sakilacustomer.org',
              'NORMAN.CURRIER@sakilacustomer.org',
              'NORMAN.CURRIER@sakilacustomer.org',
              'CORY.MEEHAN@sakilacustomer.org',
              'CORY.MEEHAN@sakilacustomer.org',
              'CORY.MEEHAN@sakilacustomer.org',
              'CORY.MEEHAN@sakilacustomer.org',
              'CORY.MEEHAN@sakilacustomer.org',
              'CORY.MEEHAN@sakilacustomer.org'],
    'address_id': [416, 416, 416, 416, 416, 416, 533, 533, 533, 533, 533, 533],
    'name': ['NORMAN CURRIER', 'NORMAN CURRIER', 'NORMAN CURRIER',
             'NORMAN CURRIER', 'NORMAN CURRIER', 'NORMAN CURRIER',
             'CORY MEEHAN', 'CORY MEEHAN', 'CORY MEEHAN', 'CORY MEEHAN',
             'CORY MEEHAN', 'CORY MEEHAN'],
    'category': ['Scifi', 'Action', 'Sports', 'Scifi', 'Family', 'Action',
                 'Documentary', 'Action', 'Sports', 'Scifi', 'Documentary',
                 'Sports']
})

df :

    customer_id                              email  address_id            name     category
0           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER        Scifi
1           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Action
2           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Sports
3           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER        Scifi
4           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Family
5           411  NORMAN.CURRIER@sakilacustomer.org         416  NORMAN CURRIER       Action
6           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN  Documentary
7           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN       Action
8           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN       Sports
9           527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN        Scifi
10          527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN  Documentary
11          527     CORY.MEEHAN@sakilacustomer.org         533     CORY MEEHAN       Sports