Attributed community search based on effective scoring function and elastic greedy method

In recent years, with the proliferation of rich attribute information available for entities in real-world networks and the increasing demand for more personalized community searches, attributed community search (ACS), an upgraded version of the community search problem, has attracted great attention from the both academic and industry areas. Some algorithms have been proposed to solve this novel research problem. However, they have a deficiency in evaluating the quality of the attributed community structure, which may mislead them and discover less valuable structures. In this paper, we make up for this defect, and propose the SFEG algorithm to better solve the ACS problem. SFEG designs a more effective scoring function to measure the quality of the discovered attributed community structure, and presents an elastic greedy optimization method to quickly maximize the function value to determine the target community with a specific meaning. The extensive experiments conducted on the attributed graph datasets with ground-truth communities show that our algorithm significantly outperforms the state-of-the-art.