rater

rater is a comparative framework for multimodal recommender systems. It was developed to facilitate the designing, comparing, and sharing of recommendation models.

Feature

• easy to use, rebuild and compare
• SOTA model
• classical model and deep model
• model has great influence in the industry
• model hsa been successfully applied by Google, Alibaba, Baidu and other well-known companies
• engineering oriented, not just experimental data validation

Data

1. ml-1m: http://files.grouplens.org/datasets/movielens/ml-1m.zip
2. delicious-2k: http://files.grouplens.org/datasets/hetrec2011/hetrec2011-delicious-2k.zip
3. lastfm-dataset-360K: http://mtg.upf.edu/static/datasets/last.fm/lastfm-dataset-360K.tar.gz
4. slashdot: http://snap.stanford.edu/data/soc-Slashdot0902.txt.gz
5. epinions: http://snap.stanford.edu/data/soc-Epinions1.txt.gz
6. ml-100k: http://files.grouplens.org/datasets/movielens/ml-100k.zip
7. Criteo(dac full): https://s3-eu-west-1.amazonaws.com/kaggle-display-advertising-challenge-dataset/dac.tar.gz
8. Criteo(dac sample): http://labs.criteo.com/wp-content/uploads/2015/04/dac_sample.tar.gz

Install

pip3 install rater

or

git clone https://github.com/shibing624/rater.git
cd rater
python3 setup.py install

Usage

Load the built-in MovieLens 1M dataset (will be downloaded if not cached):

Output:

	MAE	RMSE	AUC	NDCG@10	Recall@10	Train (s)	Test (s)
[MF]	0.7430	0.8998	0.7445	0.0479	0.0352	0.13	1.57

For more details, please take a look at our examples.

Models

The models supported are listed below. Why don't you join us to lengthen the list?

Click Through Rate Prediction

model	paper
LR: Logistic Regression	Simple and Scalable Response Prediction for Display Advertising
FM: Factorization Machine	[ICDM 2010]Factorization Machines
GBDT+LR: Gradient Boosting Tree with Logistic Regression	Practical Lessons from Predicting Clicks on Ads at Facebook
FNN: Factorization-supported Neural Network	[ECIR 2016]Deep Learning over Multi-field Categorical Data: A Case Study on User Response Prediction
PNN: Product-based Neural Network	[ICDM 2016]Product-based neural networks for user response prediction
Wide and Deep	[DLRS 2016]Wide & Deep Learning for Recommender Systems
DeepFM	[IJCAI 2017]DeepFM: A Factorization-Machine based Neural Network for CTR Prediction
AFM: Attentional Factorization Machine	[IJCAI 2017]Attentional Factorization Machines: Learning the Weight of Feature Interactions via Attention Networks
NFM: Neural Factorization Machine	[SIGIR 2017]Neural Factorization Machines for Sparse Predictive Analytics
FFM: Field-aware Factorization Machine	[RecSys 2016]Field-aware Factorization Machines for CTR Prediction
CCPM: Convolutional Click Prediction Model	[CIKM 2015]A Convolutional Click Prediction Model
Piece-wise Linear Model	[arxiv 2017]Learning Piece-wise Linear Models from Large Scale Data for Ad Click Prediction
DCN: Deep & Cross Network	[ADKDD 2017]Deep & Cross Network for Ad Click Predictions
xDeepFM	[KDD 2018]xDeepFM: Combining Explicit and Implicit Feature Interactions for Recommender Systems
AutoInt	[arxiv 2018]AutoInt: Automatic Feature Interaction Learning via Self-Attentive Neural Networks
ONN	[arxiv 2019]Operation-aware Neural Networks for User Response Prediction
FGCNN	[WWW 2019]Feature Generation by Convolutional Neural Network for Click-Through Rate Prediction
FiBiNET	[RecSys 2019]FiBiNET: Combining Feature Importance and Bilinear feature Interaction for Click-Through Rate Prediction
FLEN	[arxiv 2019]FLEN: Leveraging Field for Scalable CTR Prediction

Sequential Recommendation

model/keywords	paper
GRU4Rec	Session-based Recommendations with Recurrent Neural Networks
Caser	Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding
DIN: Deep Interest Network	[KDD 2018]Deep Interest Network for Click-Through Rate Prediction
Self-Attention	Next Item Recommendation with Self-Attention
Hierarchical Attention	Sequential Recommender System based on Hierarchical Attention Networks
DIEN: Deep Interest Evolution Network	[AAAI 2019]Deep Interest Evolution Network for Click-Through Rate Prediction
DISN: Deep Session Interest Network	[IJCAI 2019]Deep Session Interest Network for Click-Through Rate Prediction

Embedding Methods

model	paper
node2vec	node2vec: Scalable Feature Learning for Networks
item2vec	ITEM2VEC: Neural item embedding for collaborative filtering
Airbnb embedding	Real-time Personalization using Embeddings for Search Ranking at Airbnb
EGES: Enhanced Graph Embedding with Side information	Billion-scale Commodity Embedding for E-commerce Recommendation in Alibaba

Model Evolution

refer: https://zhuanlan.zhihu.com/p/63186101

Contribute

Your contributions at any level of the library are welcome. If you intend to contribute, please:

• Fork the rater repository to your own account.
• Make changes and create pull requests.

You can also post bug reports and feature requests in GitHub issues.

License

Apache License 2.0

Reference

• [Multilayer Perceptron Based Recommendation]
• [Autoencoder Based Recommendation]
• [CNN Based Recommendation]
• [RNN Based Recommendation]
• [Restricted Boltzmann Machine Based Recommendation]
• [Neural Attention Based Recommendation]
• [Neural AutoRegressive Based Recommendation]
• [Deep Reinforcement Learning for Recommendation]
• [GAN Based Recommendation]
• [Deep Hybrid Models for Recommendation]
• maciejkula/spotlight
• shenweichen/DeepCTR
• Magic-Bubble/RecommendSystemPractice
• nzc/dnn_ctr