Simple KGCN and GraphSAINT for Recommendation System

Dang Lam San - Thesis 2021.

Video demo for the training phase

Video Demo.

Step for install

Install requirements

pip install torch-scatter torch-sparse -f https://pytorch-geometric.com/whl/torch-1.8.0+cu101.html

Download data and knowledge graph from drive.

Extract data.zip to folder data
unzip ./data.zip

Download movielens-20M and extract.

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

Preprocess data

cd graphsaint/kgraphsaint
python preprocess.py -d music
python preprocess.py -d movie
cd ../../ \

Building C++ module for Sampling

python graphsaint/setup.py build_ext --inplace

Training Phase

python -m graphsaint.kgraphsaint.train --lr 1e-3 --sampler node --l2_weight 1e-5

Experimental results

Model	MovieLens – 20M	MovieLens – 20M	Last.FM	Last.FM
	AUC	F1	AUC	F1
KGCN	0.975	0.928	0.791	0.710
SKGCN – Node	0.977	0.931	0.808	0.725
SKGCN – Edge	0.977	0.977	0.811	0.731
SKGCN – Rw	0.976	0.929	0.810	0.731