Link similarity reveals multiscale link communities

Abstract

Overlapping community detection has been a hot topic in the research of complex network. In this paper, we proposed a novel link clustering method (NLC) for overlapping community detection. The method is consisted of two main steps. First step is a link similarity. The link similarity is to use a link similarity with a property of convergence to consider relationship of undirected links. The second step combines Markov Clustering Method with link similarity matrix got by first step with an extended measure of quality of modularity to determine the best partition of link communities. Extensive experiments on real world networks show our method is more reliable and reasonable than the other compared algorithms. Through varying parameters of our link similarity, our NLC method reveals multiscale link communities.