An event-based approach to overlapping community evolution by three-way decisions

In real-world social networks, there is increasing interest in tracking the evolution of groups of users. Graphs are usually investigated to represent networks, and there have been many approaches in targeted at studying these graphs. The study of the evolution of these graphs over time can provide tremendous insight on the behavior of entities, communities and the flow of information among them. However, most of existing methods use non-overlapping community structures and develop a framework of detecting the evolution of communities. In this paper, we present an event-based approach to overlapping community evolution based on three-way decisions, in order to capture and identify events from the dynamic network. In our approach, we first formalize a community as an interval set in view of the three-way decision theory. The three-way representation intuitively shows the trace of the objects that should play important roles over the period of observation. Then, evolutionary events are defined after introducing the new measures such as similarity, activity and influence. These events characterize complex behavioral patterns of individuals and communities over time. Finally, we demonstrate the suitability of the proposed approach by conducting experiments on real data extracted from the DBLP.