An effective and scalable overlapping community detection approach: Integrating social identity model and game theory

Highlights：

• SIMGT uses high-order proximity to weight and rewire the original network by which more topology information is introduced to model.

• SIMGT formulates community formation as a non-cooperative game among all nodes, in which each node is regarded as a self-interested player.

• Stochastic gradient ascent method is employed to update players’ strategies toward different communities, and we prove that the formulated game greatly resembles and matches how a potential game works (in the classical sense in game theory), indicating that the Nash equilibrium point must exist.

• Results on several synthetic and real life networks show that whatever weighting strategy we choose, SIMGT can gain better performance on community detection task. In particular, SIMGT achieves a best result when we choose the Jaccard coefficient.