Overlapping Communities Detection Based on Link Partition in Directed Networks

Abstract

Many complex systems can be described as networks to comprehend both the structure and the function. Community structure is one of the most important properties of complex networks. Detecting overlapping communities in networks have been more attention in recent years, but the most of approaches to this problem have been applied to the undirected networks. This paper presents a novel approach based on link partition to detect overlapping communities structure in directed networks. In contrast to previous researches focused on grouping nodes, our algorithm defines communities as groups of directed links rather than nodes with the purpose of nodes naturally belong to more than one community. This approach can identify a suitable number of overlapping communities without any prior knowledge about the community in directed networks. We evaluate our algorithm on a simple artificial network and several real-networks. Experimental results demonstrate that the algorithm proposed is efficient for detecting overlapping communities in directed networks. DOI: http://dx.doi.org/10.11591/telkomnika.v11i9.3304