Event recommendation in social networks based on reverse random walk and participant scale control

Abstract

With the merging of cyber world and physical world, event-based social networks have been playing an important role in promoting the spread of offline social events through online channels. Event recommendation in social networks, which is to recommend a list of upcoming events to a user according to his preference, has attracted a lot of research interests recently. In this paper, we study the event recommendation problem based on the graph theory. We first construct a heterogeneous graph to represent the interactions among different types of entities in an event-based social network. Based on the constructed graph, we propose a novel event scoring algorithm called reverse random walk with restart to obtain the user–event recommendation matrix. In practice, the participant capacity of an event may be constrained to a limited number of users. Then based on the user–event recommendation matrix, we further propose two participant scale control algorithms to coordinate unbalanced user arrangements among events. After the rearrangement, each user will be assigned a list of recommended events, which considers both local user preference and global event capacity. Experiment results on Meetup dataset show that the proposed method outperforms the state-of-art algorithms in terms of higher recommendation precision and larger recommendation coverage.