Modularity-Based Graph Clustering Using Harmony Search Algorithm

Real-world networks contain variety of meaningful information inside them that can be revealed. These networks can be biological, social, ecological and technological networks. Each of these contains specific information about their field. This information cannot be obtained with simple techniques. Various techniques and algorithms have been developed to uncover useful information from complex relationships inside the network. In this paper, to divide graphs according to modularity measure to subgraphs harmony search algorithm is used which is inspired by music improvisation. This algorithm has been tested with 5 different real-world networks. The obtained quantitative values for each network have been given in the tables. In addition the proposed algorithm, has achieved the best known modularity measure of Zachary´s Karate Club network which is commonly used in the literature and the latest subsets generated according to this modularity measure has been given at the end of section V. According to the results obtained from experiments it has been observed that HM algorithm gives faster results on solution of problem addressed in this study than most algorithms like genetic algorithm and bat algorithm. However, the proposed algorithm requires a larger size of harmony memory and more number of iterations for maximum modularity values.