Modeling Users’ Repost Behavior in Online Communities Using a Team of Learning Automata

Today's online communities play an important role in the flow of information such as news, educational contents, entertainment, and so on. Millions of users create different posts in this environment on a daily basis. Users will re-post some posts if they wish. Reposting has a significant effect on the transfer of information between users. Due to the large number of posts, users in these communities face the information overload problem. In this paper, the repost behavior of users in online communities is modeled. Firstly, effective factors have been identified in the behavior of user reposting, and then, using a reinforcement learning approach, users' repost behavior is anticipated. This reinforcement learning method is designed as a game for a team of random learning automata as a common pay-off game. To evaluate the proposed method, three large data sets have been gathered. Various scenarios have been used to evaluate the proposed method. Based on the results, randomized learning automata have great performance due to the features of the environment and online learning power.