Reference

@article{DBLP:journals/kbs/XuZPWYDZW23,
  author       = {Li Xu and
                  Jun Zeng and
                  Weile Peng and
                  Hao Wu and
                  Kun Yue and
                  Haiyan Ding and
                  Lei Zhang and
                  Xin Wang},
  title        = {Modeling and predicting user preferences with multiple item attributes
                  for sequential recommendations},
  journal      = {Knowl. Based Syst.},
  volume       = {260},
  pages        = {110174},
  year         = {2023},
  url          = {https://doi.org/10.1016/j.knosys.2022.110174},
  doi          = {10.1016/j.knosys.2022.110174},
  timestamp    = {Wed, 28 Jun 2023 14:32:09 +0200},
  biburl       = {https://dblp.org/rec/journals/kbs/XuZPWYDZW23.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

Usage

To use the code, enter the models directory and execute run_Model.py such as:

cd models/Caser
python run_Caser.py

SASRec: python main.py --dataset=ml-1m --train_dir=default --maxlen=200 --dropout_rate=0.2 --device=cuda

SSE-PT: python3 main.py --maxlen=200 --dropout_rate 0.2 --threshold_user 1 --threshold_item 1 --device=cuda

Note: Due to the different sample construction methods and experimental methods of different algorithms, we generate independent codes for each algorithm.

Requirements

• Tensorflow 1.1+
• Python 3.6+,
• numpy
• pandas

ToDo List

• More models
• Code refactoring
• Support tf.data.datasets and tf.estimator