coocurrence

[miguelgfierro]

#df = pd.DataFrame({"users": [0,1,2,0,0,1], "items":[0,1,0,0,0,0]}) #ok
#df = pd.DataFrame({"users": [0,1,2,0,0,1], "items":[1,1,0,0,0,0]}) #ok
df = pd.DataFrame({"users": [0,1,1,0,0,1], "items":[0,1,1,0,0,0]})

print(df["users"].nunique())
print(df["items"].nunique())
df

user_item_hits = sparse.coo_matrix(
            (np.repeat(1, df.shape[0]), 
             (df["users"], df["items"])
            ),
            shape=(df["users"].nunique(), df["items"].nunique()),
        ).tocsr()
user_item_hits.toarray()

item_cooccurrence = user_item_hits.transpose().dot(user_item_hits)
item_cooccurrence.toarray()