Overlapping Community Detection in Social Networks Using Cellular Learning Automata

The community detection problem in social networks has been broadly explored from different aspects. In this problem, it is attempted to divide a network into a group-set of similar nodes (communities), in which nodes inside a community structure reflect more similar functional characteristics or than that of other nodes outside the community. However, a typical characteristic of some real networks is that the presence of overlapping communities, where nodes can be a member of quite one community at an equivalent time. In this paper, an algorithm for overlapping community detection is proposed using cellular learning automata (CLA-OCD). In the CLA-OCD, a group of learning automata collaborates to find overlapping communities. The performance of the CLA-OCD is evaluated by several experiments, which is the design of both synthetic and real networks. Through experiments, the CLA-OCD achieves significant improvement in performance measures, including modularity, F-score, and normalized mutual information (NMI).