This is our implementation for the following paper:

Quanyu Dai, Xiao-Ming Wu, Lu Fan, Qimai Li, Han Liu, Xiaotong Zhang, Dan Wang, Guli Lin, Keping Yang, Personalized knowledge-aware recommendation with collaborative and attentive graph convolutional networks, Pattern Recognition, Volume 128, 2022, 108628, ISSN 0031-3203, https://doi.org/10.1016/j.patcog.2022.108628. (https://www.sciencedirect.com/science/article/pii/S0031320322001091)

Author: Quanyu Dai (quanyu.dai at connect.polyu.hk)

Knowledge graphs (KGs) are increasingly used to solve the data sparsity and cold start problems of collaborative filtering. Recently, graph neural networks (GNNs) have been applied to build KG-based recommender systems and achieved competitive performance. However, existing GNN-based methods are either limited in their ability to capture fine-grained semantics in a KG, or insufficient in effectively modeling user-item interactions. To address these issues, we propose a novel framework with collaborative and attentive graph convolutional networks for personalized knowledge-aware recommendation. Particularly, we model the user-item graph and the KG separately and simultaneously with an efficient graph convolutional network and a personalized knowledge graph attention network, where the former aims to extract informative collaborative signals, while the latter is designed to capture fine-grained semantics. Collectively, they are able to learn meaningful node representations for predicting user-item interactions. Extensive experiments on benchmark datasets demonstrate the effectiveness of the proposed method compared with state-of-the-arts.

The code has been tested running under Python 3.6.5. The required packages are as follows:

• python == 3.6.10
• tensorflow-gpu == 1.15.2
• numpy == 1.19.1
• scipy == 1.1.0
• sklearn == 0.23.1

The instruction of commands has been clearly stated in the code (see src/main.py).

• Movie

python main.py --dataset movie --aggregator 'sum' --n_epochs 20 --neighbor_sample_size 4 --dim 32 --n_iter 2 --batch_size 65536 --l2_weight 5e-6 --lr 2e-2 --layer_size [32] --adj_type plain --alg_type gcmc --model_type KGCN_NGCF --node_dropout [0.1] --mess_dropout [0.1] --node_dropout_flag 1 --agg_type weighted_avg --alpha 0 --smoothing_steps 1 --pretrain 0 --att 'u_r' --runs 3 --gpu_id 0

• book

python main.py --dataset book --aggregator 'sum' --n_epochs 20 --neighbor_sample_size 8 --dim 64 --n_iter 1 --batch_size 256 --l2_weight 2e-5 --lr 5e-5 --layer_size [64] --adj_type norm --alg_type gcn --model_type KGCN_GCN --node_dropout [0.1] --mess_dropout [0.1] --node_dropout_flag 1 --alpha 0 --smoothing_steps 3 --pretrain 0 --att 'uhrt_bi' --runs 3 --gpu_id 0

• Music

python main.py --dataset music --aggregator 'sum' --n_epochs 10 --neighbor_sample_size 8 --dim 32 --n_iter 1 --batch_size 128 --l2_weight 1e-4 --lr 0.005 --layer_size [32] --adj_type norm --alg_type gcn --model_type KGCN_GCN --node_dropout [0.1] --mess_dropout [0.1] --node_dropout_flag 1 --alpha 0.5 --smoothing_steps 8 --pretrain 0 --att 'uhrt_bi' --runs 3 --gpu_id 0

• Restaurant

python main.py --dataset restaurant --aggregator 'sum' --n_epochs 20 --neighbor_sample_size 4 --dim 8 --n_iter 2 --batch_size 65536 --l2_weight 1e-7 --lr 2e-2 --layer_size [8] --adj_type norm --alg_type gcn --model_type KGCN_GCN --node_dropout [0.1] --mess_dropout [0.1] --node_dropout_flag 1 --agg_type weighted_avg --smoothing_steps 1 --pretrain 0 --alpha 0.5 --att 'uhrt_bi' --runs 3 --gpu_id 0

We build our model based on the implementations of KGCN (https://github.com/hwwang55/KGCN) and NGCF (https://github.com/delldu/NGCF).

Hyperlink: https://pan.baidu.com/s/1As8RVt-yfA0qnl9VorfLmA
Password: h96j

If you would like to use our code, please cite:

@article{DAI2022108628,
title = {Personalized knowledge-aware recommendation with collaborative and attentive graph convolutional networks},
journal = {Pattern Recognition},
volume = {128},
pages = {108628},
year = {2022},
issn = {0031-3203},
author = {Quanyu Dai and Xiao-Ming Wu and Lu Fan and Qimai Li and Han Liu and Xiaotong Zhang and Dan Wang and Guli Lin and Keping Yang}
}