PA-GAN: Graph Attention Network for Preference-Aware Social Recommendation
Liyang Hou, Wenping Kong, Yali Gao, Yang Chen, Xiaoyong Li · Journal of Physics Conference Series · 2021
Abstract Social recommendation has been recently proposed by incorporating trust relationship to alleviate data-sparsity and cold-start problems. However, most of existing works only focus on friends different contribution to model user representation. They ignore users have different preference on items, and share different preference with friends. To address these problems, in this paper, we propose a novel Preference-Aware Graph Attention Network (PA-GAN) for trust recommendation. And we design three modules: item aggregation for user, friend-preference aggregation for user and user aggregation for item to model users’ local and global preference. Experiments on two publicly available datasets shows the proposed model PA-GAN outperforms the state-of-the-art recommendation models, and improves performance greatly.