In this repository, a number of sequential recommendation models are implemented using Python and Tensorflow. The implemented models cover common sequential recommendation algorithms (session based ). We implement the code in the paper in a concise way, including how to construct samples and training, to help readers better understand the paper's ideas.
So far, we have implemented these models, covering deep learning and traditional methods. Follow up to continue to update。
| model | paper | methods |
|---|---|---|
| AttRec | Next Item Recommendation with Self-Attention | self-attention |
| Caser | Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding | CNN |
| GRU4Rec | Session-based Recommendations with Recurrent Neural Networks | GRU |
| FPMC | Factorizing Personalized Markov Chains for Next-Basket Recommendation | MF+MC |
| TransRec | Translation-based Recommendation | MF |
| SASRec | Self-Attentive Sequential Recommendation | transfomer |
and so on.
To use the code, enter the models directory and execute run_Model.py such as:
cd models/AttRec
python run_Attrec.pyNote: Due to the different sample construction methods and experimental methods of different algorithms, we generate independent codes for each algorithm.
- Tensorflow 1.1+
- Python 3.6+,
- numpy
- pandas
- More models
- Code refactoring
- Support tf.data.datasets and tf.estimator