Discovering overlapping community structure in networks through co-clustering

Abstract

This paper presents a co-clustering based approach to discover overlapping community structure in networks that can be modeled by an undirected unweighted bipartite graph. The co-clustering algorithm uses Locality Sensitive Hashing (LSH) to cluster the nodes of the graph using Jaccard Index as the similarity measure. We employ recently proposed, weighted minwise sampling for randomized dimensionality reduction of given data set. Experimental results on benchmark data sets show that the proposed algorithm is able to capture the underlying community structure in complex networks.