A Skeleton-based Community Detection Algorithm for Directed Networks

Directionality is a key feature that represents information, energy or influence transmitting between connected objects in complex systems, community detection with consideration of such feature is a prior tool to investigate the real-word networks, in this paper we propose a novel skeleton-based community detection algorithm for directed networks. Firstly we extend the term of the edge intensity for undirected graphs to directed ones, then the skeleton chain is extracted out as a profile of the original directed network; with iterative splitting of network skeleton and the extended intensity-based modularity, disjoint communities for directed networks can be accurately retrieved. Experimental results on real and synthetic networks show higher accuracy of our algorithm than the existing methods. In addition to helping detect communities, network skeleton provides a different view with a whole new meaning into network research and can be used into many applications.