Evaluation and Customization of Community Detection Algorithms in Large Social Networks

Nowadays, social, natural, technological and information systems can be exhibited by complex networks having millions of nodes interconnected to each other. The extraction of comprehensive information from these massive networks call for computationally efficient methods. A promising approach to accomplish this task is to disintegrate the network into sub-units or communities and then using these identified communities to uncover relevant information. Thus, identifying communities in large scale networks plays a pivotal role in several scientific domains. In this paper, we extensively evaluate the functioning of two known algorithms and propose an improvement over one of them, in order to overcome its shortcomings to some extent, for optimal identification of community structure. We also present experimental results and evidences indicating that both the established algorithms, as well as our suggested approach, when applied to large social network datasets yields different results in terms of goodness and performance.