Dynamic community detection via evolutionary clustering

Dynamic community detection has been an efficient method to track the evolution of communities in dynamic social networks, which can be treated as an optimisation problem. Since evolutionary clustering (EC) was first proposed to optimise temporal data clustering, many EC-based algorithms have been developed to detect evolving community structure in dynamic social networks. However, there are two main drawbacks with existing EC based algorithms, which limit the efficiency and effectiveness of dynamic community detection: the classic operators cannot efficiently search for community structures and the general network presentation with an integer vector results in a limited search space. For this study, we first review recent literature regarding dynamic community detection using EC-based methods, and then develop two EC-based algorithms to efficiently detect evolving community structure by designing a migration operator in tandem with genetic operators and adopting a genome matrix-based representation for search space expansion. Compared with state-of-the-art baselines, our algorithms perform better in terms of clustering accuracy and temporal smoothness on both synthetic and real-world networks