/LibRecommender

Versatile End-to-End Recommender System

Primary LanguagePythonMIT LicenseMIT

LibRecommender

Build CI codecov pypi Downloads python versions Codacy Badge Code style: black platform visitors License

Overview

LibRecommender is an easy-to-use recommender system focused on end-to-end recommendation. The main features are:

  • Implemented a number of popular recommendation algorithms such as FM, DIN, LightGCN etc, see full algorithm list.
  • A hybrid recommender system, which allows user to use either collaborative-filtering or content-based features. New features can be added on the fly.
  • Low memory usage, automatically convert categorical and multi-value categorical features to sparse representation.
  • Support training for both explicit and implicit datasets, and negative sampling can be used for implicit dataset.
  • Provide end-to-end workflow, i.e. data handling / preprocessing -> model training -> evaluate -> serving.
  • Support cold-start prediction and recommendation.
  • Provide unified and friendly API for all algorithms. Easy to retrain model with new users/items.

Usage

pure collaborative-filtering example :

import numpy as np
import pandas as pd
from libreco.data import random_split, DatasetPure
from libreco.algorithms import SVDpp  # pure data, algorithm SVD++
from libreco.evaluation import evaluate

data = pd.read_csv("examples/sample_data/sample_movielens_rating.dat", sep="::",
                   names=["user", "item", "label", "time"])

# split whole data into three folds for training, evaluating and testing
train_data, eval_data, test_data = random_split(data, multi_ratios=[0.8, 0.1, 0.1])

train_data, data_info = DatasetPure.build_trainset(train_data)
eval_data = DatasetPure.build_evalset(eval_data)
test_data = DatasetPure.build_testset(test_data)
print(data_info)   # n_users: 5894, n_items: 3253, data sparsity: 0.4172 %

svdpp = SVDpp(task="rating", data_info=data_info, embed_size=16, n_epochs=3, lr=0.001,
              reg=None, batch_size=256)
# monitor metrics on eval_data during training
svdpp.fit(train_data, verbose=2, eval_data=eval_data, metrics=["rmse", "mae", "r2"])

# do final evaluation on test data
print("evaluate_result: ", evaluate(model=svdpp, data=test_data,
                                    metrics=["rmse", "mae"]))
# predict preference of user 2211 to item 110
print("prediction: ", svdpp.predict(user=2211, item=110))
# recommend 7 items for user 2211
print("recommendation: ", svdpp.recommend_user(user=2211, n_rec=7))

# cold-start prediction
print("cold prediction: ", svdpp.predict(user="ccc", item="not item",
                                         cold_start="average"))
# cold-start recommendation
print("cold recommendation: ", svdpp.recommend_user(user="are we good?",
                                                    n_rec=7,
                                                    cold_start="popular"))

include features example :

import numpy as np
import pandas as pd
from libreco.data import split_by_ratio_chrono, DatasetFeat
from libreco.algorithms import YouTubeRanking  # feat data, algorithm YouTubeRanking

data = pd.read_csv("examples/sample_data/sample_movielens_merged.csv", sep=",", header=0)
data["label"] = 1  # convert to implicit data and do negative sampling afterwards

# split into train and test data based on time
train_data, test_data = split_by_ratio_chrono(data, test_size=0.2)

# specify complete columns information
sparse_col = ["sex", "occupation", "genre1", "genre2", "genre3"]
dense_col = ["age"]
user_col = ["sex", "age", "occupation"]
item_col = ["genre1", "genre2", "genre3"]

train_data, data_info = DatasetFeat.build_trainset(
    train_data, user_col, item_col, sparse_col, dense_col
)
test_data = DatasetFeat.build_testset(test_data)
train_data.build_negative_samples(data_info)  # sample negative items for each record
test_data.build_negative_samples(data_info)
print(data_info)  # n_users: 5962, n_items: 3226, data sparsity: 0.4185 %

ytb_ranking = YouTubeRanking(task="ranking", data_info=data_info, embed_size=16,
                             n_epochs=3, lr=1e-4, batch_size=512, use_bn=True,
                             hidden_units="128,64,32")
ytb_ranking.fit(train_data, verbose=2, shuffle=True, eval_data=test_data,
                metrics=["loss", "roc_auc", "precision", "recall", "map", "ndcg"])

# predict preference of user 2211 to item 110
print("prediction: ", ytb_ranking.predict(user=2211, item=110))
# recommend 7 items for user 2211
print("recommendation(id, probability): ", ytb_ranking.recommend_user(user=2211, n_rec=7))

# cold-start prediction
print("cold prediction: ", ytb_ranking.predict(user="ccc", item="not item",
                                               cold_start="average"))
# cold-start recommendation
print("cold recommendation: ", ytb_ranking.recommend_user(user="are we good?",
                                                          n_rec=7,
                                                          cold_start="popular"))

For more examples and usages, see User Guide

Data Format

JUST normal data format, each line represents a sample. One thing is important, the model assumes that user, item, and label column index are 0, 1, and 2, respectively. You may wish to change the column order if that's not the case. Take for Example, the movielens-1m dataset:

1::1193::5::978300760
1::661::3::978302109
1::914::3::978301968
1::3408::4::978300275

Besides, if you want to use some other meta features (e.g., age, sex, category etc.), you need to tell the model which columns are [sparse_col, dense_col, user_col, item_col], which means all features must be in a same table. See above YouTubeRanking for example.

Also note that your data should not contain missing values.

Serving

For how to serve a trained model in LibRecommender, see Serving Guide .

Installation & Dependencies

From pypi :  

$ pip install LibRecommender

To build from source, you 'll first need Cython and Numpy:

$ # pip install numpy cython
$ git clone https://github.com/massquantity/LibRecommender.git
$ cd LibRecommender
$ python setup.py install

Basic Dependencies for libreco:

  • Python >= 3.6
  • TensorFlow >= 1.15
  • PyTorch >= 1.10
  • Numpy >= 1.19.5
  • Cython >= 0.29.0
  • Pandas >= 1.0.0
  • Scipy >= 1.2.1
  • scikit-learn >= 0.20.0
  • gensim >= 4.0.0
  • tqdm
  • nmslib (optional)

LibRecommender is tested under TensorFlow 1.15, 2.5, 2.8 and 2.10. If you encounter any problem during running, feel free to open an issue.

Known issue: Sometimes one may encounter errors like ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject. In this case try upgrading numpy, and version 1.22.0 or higher is probably a safe option.

The table below shows some compatible version combinations:

Python Numpy TensorFlow OS
3.6 1.19.5 1.15, 2.5 linux, windows, macos
3.7 1.20.3, 1.21.6 1.15, 2.5, 2.8, 2.10 linux, windows, macos
3.8 1.22.4, 1.23.2 2.5, 2.8, 2.10 linux, windows, macos
3.9 1.22.4, 1.23.2 2.5, 2.8, 2.10 linux, windows, macos
3.10 1.22.4, 1.23.2 2.8, 2.10 linux, windows, macos

Optional Dependencies for libserving:

Docker

One can also use the library in a docker container without installing dependencies, see Docker.

References

Algorithm Category1 Backend Sequence2 Graph3 Embedding4 Paper
userCF / itemCF pure Cython Item-Based Collaborative Filtering Recommendation Algorithms
SVD pure TensorFlow1 ✔️ Matrix Factorization Techniques for Recommender Systems
SVD++ pure TensorFlow1 ✔️ Factorization Meets the Neighborhood: a Multifaceted Collaborative Filtering Model
ALS pure Cython ✔️ 1. Matrix Completion via Alternating Least Square(ALS)
2. Collaborative Filtering for Implicit Feedback Datasets
3. Applications of the Conjugate Gradient Method for Implicit Feedback Collaborative Filtering
NCF pure TensorFlow1 Neural Collaborative Filtering
BPR pure Cython, TensorFlow1 ✔️ BPR: Bayesian Personalized Ranking from Implicit Feedback
Wide & Deep feat TensorFlow1 Wide & Deep Learning for Recommender Systems
FM feat TensorFlow1 Factorization Machines
DeepFM feat TensorFlow1 DeepFM: A Factorization-Machine based Neural Network for CTR Prediction
YouTubeRetrieval feat TensorFlow1 ✔️ ✔️ Deep Neural Networks for YouTube Recommendations
YouTubeRanking feat TensorFlow1 ✔️ Deep Neural Networks for YouTube Recommendations
AutoInt feat TensorFlow1 AutoInt: Automatic Feature Interaction Learning via Self-Attentive Neural Networks
DIN feat TensorFlow1 ✔️ Deep Interest Network for Click-Through Rate Prediction
Item2Vec pure / ✔️ ✔️ Item2Vec: Neural Item Embedding for Collaborative Filtering
RNN4Rec / GRU4Rec pure TensorFlow1 ✔️ ✔️ Session-based Recommendations with Recurrent Neural Networks
Caser pure TensorFlow1 ✔️ ✔️ Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding
WaveNet pure TensorFlow1 ✔️ ✔️ WaveNet: A Generative Model for Raw Audio
DeepWalk pure / ✔️ ✔️ DeepWalk: Online Learning of Social Representations
NGCF pure PyTorch ✔️ ✔️ Neural Graph Collaborative Filtering
LightGCN pure PyTorch ✔️ ✔️ LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation

[1] Category: pure means collaborative-filtering algorithms which only use behavior data, feat means other side-features can be included. ↩

[2] Sequence: Algorithms that leverage user behavior sequence. ↩

[3] Graph: Algorithms that leverage graph information, including Graph Embedding (GE) and Graph Neural Network (GNN) . ↩

[4] Embedding: Algorithms that can generate final user and item embeddings. ↩

Powered by

JetBrains Logo