Multiobjective discrete particle swarm optimization for community detection in dynamic networks

¹ School of Computer and Information Science, Southwest University - Chongqing 400715, China
² The Cyberspace Institute of Advanced Technology, Guangzhou University - Guangzhou 510006, China
³ School of Mechanical Engineering and Center for OPTical IMagery Analysis and Learning (OPTIMAL), Northwestern Polytechnical University - Xian 710072, China

^(a) cgao@swu.edu.cn
^(b) tianzhihong@gzhu.edu.cn
^(c) lishudong@gzhu.edu.cn
^(d) zhenwang0@gmail.com

Received: 10 April 2018
Accepted: 16 May 2018

Abstract

Tracking and identifying the dynamic patterns of evolving communities has recently drawn great attention. How to detect the community structure in a dynamic network has become a popular problem in the field of complex network and evolutionary computing. As a new concept, evolutionary clustering, is proposed to detect the process of dynamic networks under the temporal smoothness framework. Evolutionary-based clustering approaches try to maximize clustering accuracy at the current time step and minimize clustering drift at two successive time steps. But the low accuracy and the pre-setting of parameters limit their effectiveness. In order to overcome these weaknesses, in this paper, the community detection in a dynamic network is transformed into a multiobjective optimization problem. Specifically, we propose a novel decomposition strategy for multiobjective discrete particle swarm optimizationm, which balances the accuracy and the smoothness. The experimental results on synthetic and real-world datasets demonstrate the superiority of the proposed method compared with other state-of-the-art methods.