how to cumulative sum the unique items in pandas?

Question

I have a pandas df that contains date level data and

df

MONTH_YEAR    class id accnt_id 
2012-01       fruits 1 an
2012-01       fruits 2 abc
2012-01       fruits 1 def
2012-02       fruits 2 abc
2012-02       fruits 2 andi
2011-01       vege   1 an

and so on..

current query:

df.groupby(['class', 'MONTH_YEAR']).agg({'id': 'nunique', 'accnt_id': 'nunique'})

need output as:

class  MONTH_YEAR   id  accnt_id cumsum_unique_id
fruits 2012-01      2   3         3
fruits 2012-02      1   2         4
vege   2011-01      1   1         1

how to get cumsum_unique_id?

Answer 1

You need one more step to get the cumsum_unique_id

s=df.groupby(['class', 'MONTH_YEAR']).agg({'id': 'nunique', 'accnt_id': 'nunique'})

s1=df.drop_duplicates(['class','accnt_id']).\
     groupby(['class', 'MONTH_YEAR']).accnt_id.count().groupby(level=0).cumsum()
s['cumsum_unique_id']=s1
s
Out[39]: 
                   id  accnt_id  cumsum_unique_id
class  MONTH_YEAR                                
fruits 2012-01      2         3                 3
       2012-02      1         2                 4
vege   2011-01      1         1                 1