A new community detection algorithm based on fuzzy measures

Abstract

Community detection problems are one of the most important topics in social network analysis. Most of the algorithms and techniques that find communities in a network, model and represent is as something crisp. However, there exist many real situations in which fuzzy uncertainty appears in a natural way when the network is modeled. In this work, we present a modification of the well-known Louvain algorithm for crisp network that allows us to deal with fuzzy information in the network. In particular, we incorporate to the classical Louvain algorithm the use of fuzzy measures for the nodes of the graph. We also incorporate to the classical method the use of Shapley value to measure the importance of each node. We define the affinity among a pair of nodes as how each node of the pair is affected by the absence of the other one.