Social Recommendation via Graph Attentive Aggregation

• Published in: Parallel and Distributed Computing, Applications and Technologies Vol. 13148; pp. 369 - 382
• Main Authors: Liufu, Yuanwei, Shen, Hong
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2022; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Graph neural network; Parallel computing; Recommendation system; Social recommendation
• ISBN: 9783030967710; 3030967719
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-96772-7_34

Recommender systems play an important role in helping users discover items of interest from a large resource collection in various online services. Although deep graph neural network-based collaborative filtering methods have achieved promising performance in recommender systems, they are still some weaknesses. Firstly, existing graph neural network methods only take user-item interactions into account neglecting direct user-user interactions which can be obtained from social networks. Secondly, they treat the observed data uniformly without considering fine-grained differences in importance or relevance in the user-item interactions. In this paper, we propose a novel graph neural network social graph attentive aggregation (SGA) which is suitable for parallel training to boost efficiency which is the common bottleneck for neural network deployed machine learning models. This model obtains user-user collaborative information from social networks and utilizes self-attention mechanism to model the differentiation of importance in the user-item interactions. We conduct experiments on two real-world datasets and the results demonstrate that our method is effective and can be trained in parallel efficiently.