Determining Network Communities Based on Modular Density Optimization

Background: In today’s world, complex systems are conceptually observed in the form of network structure. Communities inherently existing in the networks have a recognizable elucidation in understanding the organization of networks. Community discovery in networks has grabbed the attention of researchers from multi-discipline. Community detection problem has been modeled as an optimization problem. In broad-spectrum, existing community detection algorithms have adopted modularity as the optimizing function. However, the modularity is not able to identify communities of smaller size as compared to the size of the network. Methods: This paper addresses the problem of the resolution limit posed by modularity. Modular density measure succeeds in countering the resolution limit problem. Finding network communities with maximum modular density is an NP-hard problem In this work, the discrete bat algorithm with modular density as the optimization function is recommended. Results: Experiments are conducted on three real-world datasets. For determining the consistency, ten independent runs of the proposed algorithm has been carried out. The experimental results show that our proposed algorithm produces high-quality community structure along with small size communities. Conclusion: The results are compared with traditional and evolutionary community detection algorithms. The final outcome shows the superiority of discrete bat algorithm with modular density as the optimization function with respect to number of communities, maximum modularity, and average modularity.