Multidimensional empirical analysis of overlapping community detection methods in social networks

The widespread domain of social network analysis leads to numerous research challenges associated with it. Community detection is one of the foremost research challenges. There are several community detection methods available in literature whose effectiveness for detecting communities has been analysed through evaluation of various metrics. But this criteria of empirical analysis for predicting performance of particular community detection method, needs to be further explored and refined. Major challenge with earlier surveys on empirical analysis of overlapping community detection methods is the lack of multidimensional framework for depicting the results. In literature, majority of analysis have been done by considering performance metrics only. Unlike other empirical analysis represented in literature, this paper emphasizes on analysis of interdependencies among various fitness metrics while detecting communities. Co-performance analysis based on partition comparison of overlapping community detection methods is also presented. The evaluation has been performed on real as well as benchmark datasets. This article can serve as a reference work for researchers in selection of particular overlapping community detection algorithm based on the analysis of partition comparison and inter-dependencies among fitness metrics.