Community detection in directed graphs via node similarity computation

In network analysis research domain, since a lot of object and their relations are modeled as networks or graphs, network science provides a significant tool and an indispensible platform to track their complexity. Graphs exhibit a very special property: community structure. In this paper, we propose a novel community detection method in directed graphs via node similarity computation. We focus on the community detection on directed graphs by symmetrizing the directed graphs into undirected graphs so that previous work on may subsequently be leveraged. Our main contributions include 1) we introduce two methods for computing the similarity between two nodes directed graphs; 2) test the similarity computing methods on performance of the community detection on the generated data by LFM benchmark; 3) analyze the methods based on the results of the experiments. Experiment results and analysis show the performances of two methods which we proposed are better than the previous one in the directed graphs with the significant mixing patterns.