Community detection based on strong Nash stable graph partition

The modern science of network mining has brought significant advances to the understanding of complex social networks by studying the modular structures of the networks. The problem of identifying the network modules, also called communities, is an active area of research across several disciplines due to its potential value in real-life applications. The mainstream approach for community detection generally focuses on optimization of a global metric that measures the quality of a partition over a given network. Optimizing a global metric is akin to community assignment by a centralized decision maker. However, communities in a social network often evolve in a decentralized fashion. To model the natural formation of a community, we propose a game theory-based framework treating each node as a player of a hedonic coalition formation game. We propose a novel preference relation for the players based on the players’ preference to form stable and dense community structures. The notion of strong Nash stability is used, for the first time, to determine the outcome of the coalition formation game in the context of community detection. Subsequently, we propose a greedy algorithm to obtain the strong Nash stable partition of the game. Evaluation on the well-known synthetic benchmarks and real-world networks demonstrate the superiority of the proposed algorithm.