An Effective Method for Community Search in Large Directed Attributed Graphs

Abstract

Recently there is an increasing need for online community analysis on large scale graphs. Community search (CS), which can retrieve communities efficiently on a query request, has received significant research attention. However, existing CS methods leave edge direction and vertex attributes out of consideration, which results in poor performance of community accuracy and cohesiveness. In this paper, we propose DACQ (directed attribute community query), a novel framework of retrieving effective communities in directed attributed graphs. DACQ first supplements attributes according to the topological structure and generate attribute combinations, after which DACQ finds the strongly connected k-cores (k-SCS) with attributes in the directed graph. Finally, DACQ retrieves effective communities, which are cohesive in terms of the structure and attributes. Extensive experiments demonstrate the efficiency and effectiveness of our proposed algorithms in large scale directed attributed graphs.