I want to create a recommendation system with the factorization_recommender tool that comes from GraphLab. We have user scores on certain hashtags, advertising scores (we call eclipse) on certain togs and even user scores on certain ads.
- Users scores on certain hashtags:
called df_p
count hashtag_id subscriber_id
0 22 321 172
1 44 321 161
2 25 322 172
3 6 323 172
4 32 325 172
5 26 328 161
... ... ... ...
- advertising scores on certain hastags:
called df_q
count eclipse_id hashtag_id
0 1 6521 321
1 1 6606 321
2 1 6609 321
3 1 6617 321
4 1 6649 321
5 1 6911 321
... ... ... ...
- users' scores on certain ads:
How can I use GraphLab factorization_recommender to recommend ads on subscribers?
The output should be:
subscriber_id eclipse_id score rank
13 6565 0.059975420017 1
13 6588 0.0389804676959 2
13 9762 0.0159731995118 3
13 9606 0.0159731995118 4
13 9854 0.0159731995118 5
13 9576 0.0159731995118 6
13 9902 0.0155536116738 7
13 9875 0.0155536116738 8
13 6766 0.0126994707082 9
13 9870 0.0125380719963 10
About GraphLab I have:
user_info = graphlab.SFrame({'user_id': ["0", "1", "2"],
'name': ["Alice", "Bob", "Charlie"],
'numeric_feature': [0.1, 12, 22]})
item_info = graphlab.SFrame({'item_id': ["a", "b", "c", d"],
'name': ["item1", "item2", "item3", "item4"],
'dict_feature': [{'a' : 23}, {'a' : 13},
{'b' : 1},
{'a' : 23, 'b' : 32}]})
m2 = graphlab.factorization_recommender.create(sf, target='rating',
user_data=user_info,
item_data=item_info)
But it seems that does not apply to my case so far we have scores on users and scores on item.