Mining Target Attribute Subspace and Set of Target Communities in Large Attributed Networks

Abstract: Community detection provides invaluable help for various applications, such as marketing and product recommendation. Traditional community detection methods designed for plain networks may not be able to detect communities with homogeneous attributes inside on attributed networks with attribute information. Most of recent attribute community detection methods may fail to capture the requirements of a specific application and not be able to mine the set of required communities for a specific application. In this paper, we aim to detect the set of target communities in the target subspace which has some focus attributes with large importance weights satisfying the requirements of a specific application. In order to improve the university of the problem, we address the problem in an extreme case where only two sample nodes in any potential target community are provided. A Target Subspace and Communities Mining (TSCM) method is proposed. In TSCM, a sample information extension method is designed to extend the two sample nodes to a set of exemplar nodes from which the target subspace is inferred. Then the set of target communities are located and mined based on the target subspace. Experiments on synthetic datasets demonstrate the effectiveness and efficiency of our method and applications on real-world datasets show its application values.