Empirical Analysis and Modeling of Users' Topic Interests in Online Forums

Bulletin Board Systems (BBSs) have demonstrated their usefulness in spreading information. In BBS forums, a few posts that address currently popular social topics attract a lot of attention, and different users are interested in many different discussion topics. We investigate topic cluster features and user interests of an actual BBS forum, analyzing user posting and replying behavior. According to the growing process of BBS, we suggest a network model in which each agent only replies to the posts that belong to its specific topics of interest. A post that is replied to will be immediately assigned the highest priority on the post list. Simulation results show that characteristics of our model are similar to those of the real BBS. The model with heterogeneous user interests promotes the occurrence of popular posts, and the user relationship network possesses a large clustering coefficient. Bursts and long waiting time exist in user replying behavior, leading to non-Poisson user activity pattern. In addition, the model produces an analogous evolving trend of Gini coefficients for posts' and clusters' participants as BBS forums.