An Extended Louvain Method for Community Detection in Attributed Social Networks

Community detection is a significant way to analyze complex networks. Classical methods usually deal only with the network's structure and ignore content features. During the last decade, most solutions for community detection only consider network topology. Social networks, as complex systems, contain actors with certain social connections. Moreover, most real-world social networks provide additional data about actors, such as age, gender, preferences, etc. However, content-based methods lead to the loss of valuable topology information. This paper describes and clarifies the problems and proposes a fast and deterministic method for discovering communities in social networks to combine structure and semantics. The proposed method has been evaluated through simulation experiments, showing efficient performance in network topology and semantic criteria and achieving proportional performance for community detection.