This is the Pytorch implementation of KGCN (arXiv):

Knowledge Graph Convolutional Networks for Recommender Systems
Hongwei Wang, Miao Zhao, Xing Xie, Wenjie Li, Minyi Guo.
In Proceedings of The 2019 Web Conference (WWW 2019)

• Movie

  Raw rating file for movie is too large to be contained in this repo.

  Downlad the rating data first

  $ wget http://files.grouplens.org/datasets/movielens/ml-20m.zip
  $ unzip ml-20m.zip
  $ mv ml-20m/ratings.csv data/movie/

• Music

  Nothing to do

• Other dataset

  If you want to use your own dataset, you need to prepare 2 data.

  1. Rating data
     • Each row should contain (user-item-rating)
     • In this repo, it is pandas dataframe structure. (look at data_loader.py)
  2. Knowledge graph
     • Each triple(head-relation-tail) consists of knowledge graph
     • In this repo, it is dictionary type. (look at data_loader.py)

1. data_loader.py

   • data loader class for movie / music dataset
   • you don't need it if you make custom dataset

2. aggregator.py

   • aggregator class which implements 3 aggregation functions

3. model.py

   • KGCN model network

Look at the KGCN.ipynb.

It contains

• how to construct Datset
• how to construct Data loader
• how to train network