Fatigue-Aware Event-Participant Arrangement in Event-Based Social Networks: An Upper Confidence Bound Method

Abstract

As the mobile Internet and social computing developed, online event-based social networks (EBSNs) were derived, which mainly assign events to users according to the scores a linear combination of some features (i.e., location, similarity, friendship). Most of existing research work only takes offline scenarios into consideration, where users’ full information is known in advance. However, on real-world EBSN platforms, online scenarios have practical application value. Besides, existing works did not consider online learning users’ feedbacks (i.e., accept or reject arrangement), and did not consider users’ fatigue from seeing less interest events. In this paper, we investigate the online users’ feedback and fatigue control, where users can give a feedback by accepting a set of events arranged or reject events arranged due to less interesting events. In particular, we first model the problem as a stochastic bandit, and then applying Upper Confidence Bound based method (UCB-based) with expected regret, which is the polynomial in the events quantity in combinatorial settings. Finally, we evaluate the performance of our proposed algorithms with real data sets and synthetic data sets. And we find that in most setting, UCB-based algorithm has much lower regret compared to current state-of-the-art technology.