Revealing network communities through modularity maximization by a contraction–dilation method

Abstract

Many real-world systems can be described by networks whose structures relate to functional properties. An important way to reveal topology–function correlations is to detect the community structures, which can be well evaluated by graph modularity. By maximizing modularity, large networks can be divided into naturally separated groups. Here, we propose a contraction–dilation algorithm based on single-node-move operations and a perturbation strategy. Tests on artificial and real-world networks show that the algorithm is efficient for discovering community structures with high modularity scores and accuracies at low expenses of both time and memory.