A parallel self-organizing overlapping community detection algorithm based on swarm intelligence for large scale complex networks

Abstract

Community detection is a critical task for complex network analysis. It helps us to understand the properties of the system that a complex network represents and has significance to a wide range of applications. Though a large number of algorithms have been developed, the detection of overlapping communities from large scale and (or) dynamic networks still remains challenging. In this paper, a Parallel Self-organizing Overlapping Community Detection (PSOCD) algorithm ground on the idea of swarm intelligence is proposed. The PSOCD is designed based on the concept of swarm intelligence system where an analyzed network is treated as a decentralized, self-organized, and self-evolving systems, in which each vertex acts iteratively to join to or leave from communities based on a set of predefined simple vertex action rules. The algorithm is implemented on a distributed graph processing platform named Giraph++; therefore it is capable of analyzing large scale networks. The algorithm is also able to handle overlapping community detection well because a vertex can naturally joins to multiple communities simultaneously. Moreover, if some vertexes and edges are added to or deleted from the analyzed network, the algorithm only needs to adjust community assignments of affected vertexes in the same way as its finding joining communities for a vertex, i.e., it inherently supports dynamic network analysis. The proposed PSOCD is evaluated using a number of variety large scale synthesized and real world networks. Experimental results indicate that the proposed algorithm can effectively discover overlapping communities on large-scale network and the quality of its detected overlapping community structures is superior to two state-of-the-art algorithms, namely Speaker Listener Label Propagation Algorithm (SLPA) and Order Statistics Local Optimization Method (OSLOM), especially on high overlapping density networks and (or) high overlapping diversity networks.