This is the official implementation of PET (Personalized View Weighting with Data Enhancement Two-Pronged Contrast) (CIKM 2024 Short Paper Track)

For additional data anlysis results and details about loss function, you can check here. Supplementary Document

)

We use three widely used datasets for bundle recommendation, iFashion, NetEase and Youshu. For the iFashion dataset, please unzip data.zip in the same folder.

cd PET

• iFashion

python train.py -d iFashion -g [gpu_id]

• NetEase

python train.py -d NetEase -g [gpu_id]

• Youshu

python train.py -d Youshu -g [gpu_id]   

We appreciate your interest in our work. If our research contributes to your projects, please consider citing our paper:

@inproceedings{kim2024towards,
  title={Towards Better Utilization of Multiple Views for Bundle Recommendation},
  author={Kim, Kyungho and Kim, Sunwoo and Lee, Geon and Shin, Kijung},
  booktitle={CIKM},
  year={2024}
}

This code is implemented based on the open source code from the paper CrossCBR : Cross-view Contrastive Learning for Bundle Recommendation (KDD '22).