Community Detection Using Maximizing Modularity and Similarity Measures in Social Networks

Abstract

We introduce a new method to unveil the structure of community in large complex networks. The proposed community detection algorithm uses the well-known notion of network modularity optimization. To achieve this, our method uses a cosine similarity measure which depends on shared links. This similarity measure helps to efficiently find the similarity between nodes in large networks. It also considers the sparse nodes present in the network with low complexity as compared to other similarity measures. Once the similarity is computed, the method selects the pairwise proximity among the nodes of the complex networks. Further, it detects the communities taking a procedure motivated from well-known state-of-the-art Louvain method efficiently maximizing the modularity value. We carried out experiments to show that our method surpasses other approaches and marginally improve the results of the other existing methods, providing reliable results. The performance analysis of methods evaluated in terms of communities, modularity value, and quality of community obtained in the network.