An evolutionary dynamic optimization framework for structure change detection of streaming networks

Community detection in complex networks that evolve over time is a challenging task, intensively studied in the last few years. An algorithm designed for this problem should be able to successfully discover if changes occurred in the network, and quickly react by modifying the community structure for reflecting the new network organization. Evolutionary dynamic optimization is a powerful technique to solve time-dependent problems by applying evolutionary algorithms. In this paper we propose to exploit it for time evolving graphs in order to discover changes in community structure and to quickly adapt when such changes occur. The approach uses a population-based model for change detection, and applies two different strategies to adapt to changes. Experimental results on synthetic networks show the very good performances of evolutionary dynamic optimization to deal with this kind of problem.