CODEWITHSUNDEEP
pythonpandasgroup-byapply
Black
Black

Reputation: 4654

Python: Apply function to groupby

I have some dataframe in the form

userid |      event_time      | activity
   A      2017-01-01 02:20:34      E1
   A      2017-01-01 02:20:50      E2     
   A      2017-03-01 11:23:43      E1                
   A      2017-03-01 11:23:55      E6
   B      2017-01-01 08:24:32      E1                
   B      2017-01-01 08:25:25      E4
   C      2017-01-01 23:19:21      E3                
   C      2017-01-01 23:20:04      E11

I would like to apply a function to each group (grouped by userid). That counts the number of times a user has re-experienced the same event that they had encountered. For example, user A has re-experienced E1 at 2017-03-01 11:23:43.

 userid   |  activity | cnt_previous_events
    A           E1             0
    A           E2             0
    A           E1             1   
    A           E6             0

I have tried the following:

def previous_event_ctr(group):

    events = set()
    ctr = 0

    for val in group:
        if val in events:
           ctr += 1
        else:
           events.add(val)

    return ctr

And applied the following to my dataframe column, 

df.groupby('userid').activity.agg(previous_event_ctr)

But I keep getting a TypeError: 'Series' objects are mutable, thus they cannot be hashed. How should I be applying this the function to my dataframe using groupby.

Upvotes: 3

Views: 459

Answers (1)

jezrael
jezrael

Reputation: 863531

It seems you need cumcount, df has to be sorted by userid and event_time first:

df['count'] = df.sort_values(['userid','event_time'])
                .groupby(['userid', 'activity']).activity.cumcount()
print (df)

  userid           event_time activity  count
0      A  2017-01-01 02:20:34       E1      0
1      A  2017-01-01 02:20:50       E2      0
2      A  2017-03-01 11:23:43       E1      1
3      A  2017-03-01 11:23:55       E6      0
4      B  2017-01-01 08:24:32       E1      0
5      B  2017-01-01 08:25:25       E4      0
6      C  2017-01-01 23:19:21       E3      0
7      C  2017-01-01 23:20:04      E11      0

Upvotes: 5

Related Questions