Community detection in heterogenous networks using incremental seed expansion

Abstract

Mining communities is fundamental to network science as it facilitates the analysis of modern information infrastructure. Though community detection in networks is a very active research area, most of the algorithms focus on homogenous networks. Homogeneous networks consist of only one type of nodes and one type of edges. However, real-world networks mostly exhibit multi-typed objects and relationships thus forming heterogeneous information networks. There have been only a few studies in the past which aim at revealing the hidden communities in heterogeneous information networks. In this paper, we propose an effective community detection algorithm based on incremental seed expansion mechanism. Our algorithm utilizes the concept of weighted path matrix and transforms the network from heterogeneous to a homogenous network. The central idea behind our algorithm is to identify good seeds and then iteratively expand them such that the conductance score of detected communities is minimized. Experiment on a real-world network establish the viability of proposed approach.