Detection of dynamic and overlapping communities in social networks

摘要

Community detection has received much attention in recent years. However, most of community detection methods consider the network as a static structure where communities cannot overlap. A more reasonable description seems to be dynamic and overlapping community structure. Moreover, community detection using members attributes has rarely been investigated in dynamic networks. This paper proposes an incremental approach for the identification of dynamic and overlapping communities. The approach adopted takes advantage of the strengths of multi-agent systems, namely robustness, evolving and adaptive system, for an efficient community detection in dynamic social networks. We use a set of agents associated with detected communities. These agents observe the network, and consequently update their communities. To improve the accuracy of community detection, members attributes are combined with network structure information. The experimental results have shown that our method detects communities with high quality, and successfully tracks their evolution over time.