Simple example of building recommendations with RecTools

• Building simple model

• Visual recommendations checking

import numpy as np
import pandas as pd

from implicit.nearest_neighbours import TFIDFRecommender

from rectools import Columns
from rectools.dataset import Dataset
from rectools.models import ImplicitItemKNNWrapperModel

Load data

%%time
!wget -q https://files.grouplens.org/datasets/movielens/ml-1m.zip -O ml-1m.zip
!unzip -o ml-1m.zip
!rm ml-1m.zip

Archive:  ml-1m.zip
  inflating: ml-1m/movies.dat
  inflating: ml-1m/ratings.dat
  inflating: ml-1m/README
  inflating: ml-1m/users.dat
CPU times: user 134 ms, sys: 415 ms, total: 548 ms
Wall time: 4.39 s

%%time
ratings = pd.read_csv(
    "ml-1m/ratings.dat",
    sep="::",
    engine="python",  # Because of 2-chars separators
    header=None,
    names=[Columns.User, Columns.Item, Columns.Weight, Columns.Datetime],
)
print(ratings.shape)
ratings.head()

(1000209, 4)
CPU times: user 5.76 s, sys: 409 ms, total: 6.17 s
Wall time: 6.16 s

	user_id	item_id	weight	datetime
0	1	1193	5	978300760
1	1	661	3	978302109
2	1	914	3	978301968
3	1	3408	4	978300275
4	1	2355	5	978824291

%%time
movies = pd.read_csv(
    "ml-1m/movies.dat",
    sep="::",
    engine="python",  # Because of 2-chars separators
    header=None,
    names=[Columns.Item, "title", "genres"],
    encoding_errors="ignore",
)
print(movies.shape)
movies.head()

(3883, 3)
CPU times: user 9.55 ms, sys: 1.62 ms, total: 11.2 ms
Wall time: 10.4 ms

	item_id	title	genres
0	1	Toy Story (1995)	Animation|Children's|Comedy
1	2	Jumanji (1995)	Adventure|Children's|Fantasy
2	3	Grumpier Old Men (1995)	Comedy|Romance
3	4	Waiting to Exhale (1995)	Comedy|Drama
4	5	Father of the Bride Part II (1995)	Comedy

Build model

# Prepare a dataset to build a model
dataset = Dataset.construct(ratings)

%%time
# Fit model and generate recommendations for all users
model = ImplicitItemKNNWrapperModel(TFIDFRecommender(K=10))
model.fit(dataset)
recos = model.recommend(
    users=ratings[Columns.User].unique(),
    dataset=dataset,
    k=10,
    filter_viewed=True,
)

CPU times: user 6.05 s, sys: 274 ms, total: 6.32 s
Wall time: 1.42 s

# Sample of recommendations - it's sorted by relevance (= rank) for each user
recos.head()

	user_id	item_id	score	rank
0	1	364	20.436578	1
1	1	1196	15.716834	2
2	1	318	15.625371	3
3	1	2096	14.876911	4
4	1	2571	12.718620	5

Check recommendations

# Select random user, see history of views and reco for this user
user_id = 3883
user_viewed = ratings.query("user_id == @user_id").merge(movies, on="item_id")
user_recos = recos.query("user_id == @user_id").merge(movies, on="item_id")

# History, but only films that user likes
user_viewed.query("weight > 3")

	user_id	item_id	weight	datetime	title	genres
0	3883	2997	5	967134212	Being John Malkovich (1999)	Comedy
2	3883	1265	5	967134285	Groundhog Day (1993)	Comedy|Romance
4	3883	2858	5	965822230	American Beauty (1999)	Comedy|Drama
10	3883	2369	4	965822136	Desperately Seeking Susan (1985)	Comedy|Romance
14	3883	3189	4	965822296	My Dog Skip (1999)	Comedy
16	3883	1784	4	965822136	As Good As It Gets (1997)	Comedy|Drama
17	3883	2599	4	967134250	Election (1999)	Comedy
18	3883	34	4	967134285	Babe (1995)	Children's|Comedy|Drama

# Recommendations
user_recos.sort_values("rank")

	user_id	item_id	score	rank	title	genres
0	3883	2396	13.991358	1	Shakespeare in Love (1998)	Comedy|Romance
1	3883	2762	10.249648	2	Sixth Sense, The (1999)	Thriller
2	3883	318	7.728188	3	Shawshank Redemption, The (1994)	Drama
3	3883	608	7.617913	4	Fargo (1996)	Crime|Drama|Thriller
4	3883	356	5.674010	5	Forrest Gump (1994)	Comedy|Romance|War
5	3883	2395	5.508895	6	Rushmore (1998)	Comedy
6	3883	223	5.398012	7	Clerks (1994)	Comedy
7	3883	593	5.335058	8	Silence of the Lambs, The (1991)	Drama|Thriller
8	3883	296	4.828189	9	Pulp Fiction (1994)	Crime|Drama
9	3883	2959	4.615653	10	Fight Club (1999)	Drama

Here is the simple example, we only used ratings to train the model and we only prepared recommendations for users who have rated movies before. But some models allow you to use explicit features, e.g. user age or item genre. And some models allow you to generate recommendations for users that have not rated any movies before. See documentation for the details.