This repository provides PyTorch implementation of TemproxRec from the paper "Sequential Recommendation on Temporal Proximities with Contrastive Learning and Self-Attention"

We test code with python 3.11.3 on ubuntu with cuda 11.4

Install the required packages into your python environment:

pip install -r requirements.txt

The trained networks representing the best performances from our experiments for all datasets are available on our anonymous Google Drive

Their training scripts with optimal parameters are described in the 'run.sh'.

To reproduce the experiments, download all the folders and add them to the ./experiments/. Then, test the trained networks.

Training TemproxRec is produced by using the main.py. For example, to train on Amazon 'beauty'

python main.py --dataset_code='beauty' --clip_time=128 --interval=60 --temperature=0.05 --lamb=0.3 --mode=train

The running scripts for the best experiment for all datasets are described in the 'run.sh'

By changing the 'mode' to 'test_only' on the training code, the best model stored in the training stage can be tested.

python main.py --dataset_code='beauty' --clip_time=128 --interval=60 --temperature=0.05 --lamb=0.3 --mode=test_only

We experimeted with four dataset : Amazon Beauty, Video, Books, and Steam.

Preprocessed data for all are stored in ./Data/preprocessed. You can run the above training and test code with the preprocessed data.

Law data for Beauty and Video datasets are stored in ./Data/beauty and ./Data/video.

Due to capacity issues, the law data for Books and Steam datasets are available on our anonymous Google Drive.

Download the folder "books" and "steam" from the drive and add them to the TemproxRec's code in ./Data/.