Fast and accurate community search algorithm for attributed graphs

Abstract

The community search algorithm is an essential graph data management tool to identify a community suited to a user-specified query node. Although the community search algorithms are useful in various applications, it is difficult for them to handle attributed graphs since (1) traditional algorithms ignore node attributes and (2) algorithms require strict topological constraints to find a community. In this paper, we define a novel class of the community search problem on attributed graphs called the flexible attributed truss community (F-ATC) problem. To overcome the aforementioned limitations, the F-ATC problem relaxes the topological constraints and evaluates node attributes. Since the F-ATC problem is NP-hard, we propose two greedy algorithms to solve it efficiently. Our extensive experiments on real-world graphs clarify that our approach achieves higher efficiency and accuracy than the state-of-the-art method.