Content-Based Co-Factorization Machines: Modeling User Decisions in Event-Based Social Networks

Event-based online social networks (EBSNs) have attracted millions of users to attend events and join event groups. However, the EBSNs are often overwhelmed with too many events and groups, making it is hard for users to attend events and join groups that interest them. Thus, it is natural to design recommender systems to recommend events and groups to users. One key challenge is that, though users have different kinds of behaviors (e.g., user-event behavior, user-word review behavior, and user-group behavior), these data are very sparse for prediction. To that end, in this paper, we propose a content-based co-factorization machines based method for the two recommendation tasks by co-relating users' different kinds of behaviors. Besides, to alleviate the data sparsity issue, we also model the content information in the co-factorization machines. Finally, experiments on three real-world datasets show the effectiveness of our proposed model on the two prediction tasks.