Train and test a recommender

Prepare the dataset as introduced before:

[1]:

import rsdiv as rs

loader = rs.MovieLens1MDownLoader()
ratings = loader.read_ratings()
items = loader.read_items()

rsdiv provides various implementations of core recommender algorithms.

To start with, a wrapper for LightFM is also supported:

[2]:

rc = rs.FMRecommender(ratings, items).fit()

30% of interactions are split for the test set by default, and the precision at the top 5 can be calculated with:

[3]:

rc.precision_at_top_k(5)

[3]:

0.15970199

the top 100 unseen recommended items for an arbitrary user, say userId: 1024, can be simply given by:

[4]:

rc.predict_top_n_item(1024, 100)

[4]:

	itemId	scores	title	genres	release_date
0	2557	2.180831	I Stand Alone (Seul contre tous)	[Drama]	1998
1	2865	2.078251	Sugar Town	[Comedy]	1999
2	2514	2.065647	Pet Sematary II	[Horror]	1992
3	2634	1.862742	Mummy	[Horror]	1959
4	3542	1.828745	Coming Apart	[Drama]	1969
...	...	...	...	...	...
95	3582	0.799194	Jails	[Drama]	2000
96	3075	0.780045	Repulsion	[Thriller]	1965
97	2943	0.779874	Indochine	[Drama, Romance]	1992
98	2140	0.774891	Dark Crystal	[Children's, Fantasy, Sci-Fi]	1982
99	393	0.773604	Street Fighter	[Action]	1994

100 rows × 5 columns