Enhancing modularity optimization via local close-knit structures

Abstract

Discovering communities is crucial for studying the structure and dynamics of networks. Groups of related nodes in the community often correspond to functional subunits such as protein complexes or social spheres. The modularity optimization method is typically an effective algorithm with global objective function. In this paper, we attempt to further enhance the quality of modularity optimization by mining local close-knit structures. First, both periphery and core close-knit structures are defined, and several fast mining and merging algorithms are presented. Second, a novel Fast Newman (FN) algorithm named NFN incorporating local structures into global optimization is proposed. Experimental results in terms of both internal and external on six real-world social networks have demonstrated the effectiveness of NFN on community detection.