Who will be participating next?
Xuning Tang, Christopher C. Yang, Mi Zhang · 2012
Predicting whether a user will be participating in a thread has broad applications, such as thread recommendation and ranking. In an extremist forum, knowing which user will be interested to join a particular thread with sensitive or threatening information is also important for security agent to prevent or prepare for any potential outbreak of crisis. Traditional methods employed a bipartite graph to represent user-thread relationships and predict potential users for a new coming thread based on user similarities. In this paper, we propose a User Interest and Topic Detection model to extract topics and trends from a document corpus and also discover users' interests toward these trends. Information of user interest is then used to predict potential information consumers for a given thread. Experiments conducted in the Dark Web dataset showed the effectiveness of our approach; especially when we have limited information about who have already participated in an existing new thread.