KTPGN: Novel event-based group recommendation method considering implicit social trust and knowledge propagation

Abstract

Groups where like-minded people gather to share interests, comments, or participate in activities have recently gained popularity in well-known social platforms, such as Meetup and Douban. Group recommendations are essential in social media to help users identify their preferred groups. Our concern is event-based group recommendations, based on which users join a group primarily to participate in offline events hosted by group members. Traditional methods for group recommendation do not work well in addressing this issue, because they ignore the social relationships among users or multiple attributes of groups. The knowledge-enhanced trust propagation graph neural network (KTPGN) presented in this study exploits social trust, multiple attributes of groups, and user-group interaction history in a unified framework to predict user preferences for groups. First, an implicit social trust graph is constructed by considering two aspects of credibility, because explicit trust relationships are not always available in most real recommender systems. Furthermore, a graph neural network (GNN)-based model is proposed to learn group embeddings by iteratively aggregating multiple linked attributes and user embeddings through recursive propagation of social trust and historical interest. Experimental results on several datasets from Meetup demonstrate that KTPGN significantly outperforms state-of-the-art recommendation methods.