AC2CD: An actor–critic architecture for community detection in dynamic social networks

Abstract

A vital problem tackled in the network analysis literature is community structure identification in social networks. There are many solutions to the community detection problem. Nevertheless, most of the approaches are constrained to static scenarios. Some solutions adapted to dynamic social networks present performance limitations, and others do not fit well in such contexts. This situation aggravates when considering the demand to analyze constantly growing networks as real-world social networks. This work presents the Actor–Critic for Community Detection (AC2CD) architecture for community detection in dynamic social networks. The architecture based on the deep reinforcement learning strategy deals with changing aspects of large networks using a local optimization of the modularity density function. The experiments using real-world dynamic network datasets present better results than state-of-art solutions, indicating that AC2CD copes well with dynamic real-world social networks.