Groupby计算Set中唯一值的频率

Question

I have a DF like this

   User      Dept     
    1        Cook
    1        Cook
    1        Home
    2        Sports
    2        Travel
    2        Cook

I want to count the unique users within each department:

   Dept      User
   Cook      2
   Home      1
   Sports    1
   Travel    1

Notice how the department Cook only has a count of two because even though three users were found in 'Cook', there were only two unique users

I have tried the following:

 df.groupby(['Dept']).count()  -- counts 'Cook' three times
 df.drop_duplicates(['Dept']).groupby('Dept')['User'].sum() -- over counts all departments 

I know the answer is a groupby, I just can't seem to figure it out!

Answer 1

You could use nunique :

>>> df.groupby("Dept")["User"].nunique()
Dept
Cook      2
Home      1
Sports    1
Travel    1
Name: User, dtype: int64
>>> df.groupby("Dept")["User"].nunique().reset_index()
     Dept  User
0    Cook     2
1    Home     1
2  Sports     1
3  Travel     1

(Note that I used your example data, which only has one unique user in Sports.)