A Novel Approach for Community Detection Using the Label Propagation Technique

Abstract

In this chapter, we propose a new label propagation-based approach to detect community structure in social networks. This is a multiple label propagation technique in which a node can obtain labels of different communities, which allows researchers to discover overlapping communities. One important advantage of this approach is the updating of node labels over time that makes it dynamic. Given an underlying social network, we assume that each node receives a unique label id similar to its node id in the initial phase. We allow each node to accept multiple labels from its neighbors if each of the neighbors has a high common neighbor score, which naturally encompasses the idea of overlapping communities.