GCNext

Unsupervised Graph Embeddings for Session-based Recommendation with Item Features.

Use pipenv to create virtual environment.

Diginetica & Tmall datasets can be found under: https://drive.google.com/drive/folders/1ritDnO_Zc6DFEU6UND9C8VCisT0ETVp5

Copy raw datasets to data folder.

Usage

Tmall dataset as example:

• Preprocessing: in dataset folder, run:

  python tmall.py
  

• Generate GNN embeddings: in gnn folder, run:

  python bgrl.py --config ./config/tmall/bgrl.yaml 
  

• Run sequential model training: in seq_rec folder, run:

  python run.py --config config/tmall/gru4recgnn.yaml
  

Change config paths for different datasets and models.

Requirements

See Pipfile.

Citation

@inproceedings{peintner:cars:2022,
 author = {Andreas Peintner, Marta Moscati, Emilia Parada-Cabaleiro, Markus Schedl, Eva Zangerle},
 booktitle = {CARS: Workshop on Context-Aware Recommender Systems (RecSys ’22)},
 title = {Unsupervised Graph Embeddings for Session-based Recommendation with Item Features},
 year = {2022}
}

References

• https://github.com/nerdslab/bgrl
• https://github.com/rn5l/session-rec
• https://github.com/RUCAIBox/RecBole