NeuralTMT

The code and dataset for our NN 2022 paper: Factorizing Time-heterogeneous Markov Transition for Temporal Recommendation(https://doi.org/10.1016/j.neunet.2022.11.032). We have implemented our methods in Pytorch.

Dependencies

• Python 3.8
• torch 1.10.1

Usage

Generate data

You need to run the file pre_data.py to generate the data format needed for our model. For example:

u1:[u1,[i_1,i_2],[i_3,i_4,i_5],[i_3,i_8],[i_0,i_5],[i_1,i_2],[i_4,i_5],[i_6,i_7],[i_1,i_2]]
u2:[u2,[i_1,i_4],[i_5,i_2],[i_7,i_9],[i_10,i_21],[i_1,i_2],[i_3,i_4,i_5],[i_6,i_7],[i_1,i_2]]
...

Training and Testing

Then you can run the file main.py to train and test our model.

Cite

If you want to use our codes in your research, please cite:

@ARTICLE
  article{wen2022factorizing,
  title={Factorizing time-heterogeneous Markov transition for temporal recommendation},
  author={Wen, Wen and Wang, Wencui and Hao, Zhifeng and Cai, Ruichu},
  journal={Neural Networks},
  year={2022},
  publisher={Elsevier}
}