YahongANN/IDGCN

The source code of the paper "IDGCN: Individual aware Diversified Graph Convolutional Network for Recommendation"

Environment: python = 3.6 pandas = 1.1.5 torch = 1.9.1+cu111 numpy = 1.19.5 tqdm = 4.63.0

Directory:

IDGCN

• config config file

• data

  Beauty, download from https://cseweb.ucsd.edu/~jmcauley/datasets.html#amazon_reviews

  ml-10m, download from https://grouplens.org/datasets/movielens/10m/

  Music, download from http://www.cp.jku.at/datasets/Music4All-Onion

• src

  • models
    • IDGCN.py: proposed model
    • LightGCN.py: LightGCN model
    • MF.py: mf mode
  • trainer
    • base_trainer.py: trainer module
    • idgcn_trainer.py: idgcn_trainer module
  • util
    • spmm.py: sparse matrix multiplication function
  • metrics.py: some metric functions like recall, ndcg, and our proposed IHC
  • data_generator.py: date generator module

• main_mf.py: main file for conducting MF model

• main_idgcn.py : main file for conducting IDGCN model

• main_lightgcn.py:main file for conducting LightGCN model

Run the Codes: python main_mf.py or python main_ligtgcn.py or python main_idgcn.py

you can change hyperparameters by resetting the config file.

other baseline methods can be found on the link:

DivMF https://github.com/snudatalab/DivMF.

ALGCN https://github.com/AllminerLab/Code-for-ALGCN-master

DGCN https://github.com/tsinghua-fib-lab/DGCN.

DGRec https://github.com/YangLiangwei/DGRec.

Preprocessing the Beauty dataset can reference: SIGIR 2021 Enhancing Domain-Level and User-Level Adaptivity in Diversified Recommendation https://github.com/NLPWM-WHU/EDUA/blob/main/code/preprocess_beauty.py