I have a table with sessions and want to count the session that the userID goes through.

1 session and 1 count of sessions.

output is

So the count is based on a unique user with the session that the user id has been exposed to.

CodePudding user response：

IIUC, you want to find the unique combinations of sites, without duplicates (but it remains unclear whether order is important).

In order (A->B is different than B->A):

(df.groupby('userID')['session']
   .agg(lambda x: tuple(dict.fromkeys(x)))
   .value_counts()
)

output:

(facebook, google)            1
(facebook, google, tiktok)    1
(tiktok,)                     1
Name: session, dtype: int64

or, to illustrate the order:

(df.groupby('userID')['session']
   .agg(lambda x: '->'.join(dict.fromkeys(x)))
   .value_counts()
)

output:

facebook->google            1
facebook->google->tiktok    1
tiktok                      1
Name: session, dtype: int64

Order independent (A->B would count as B->A):

(df.groupby('userID')['session']
   .agg(frozenset)
   .value_counts()
)

output:

(facebook, google)            1
(facebook, tiktok, google)    1
(tiktok)                      1
Name: session, dtype: int64