Multiple-kernel combination fuzzy clustering for community detection

Abstract

Community detection on social network is a challenging task, and the multiple-kernel learning method is gaining popularity. In this paper, we propose a new multiple-kernel combination algorithm for community partitioning. We study several base kernel matrices from the adjacency matrix of a network. By adjusting the weights of different base kernel matrices, a new kernel matrix is constructed using linear combination of those matrices. To partition networks whose number of communities are known in advance, we derive a new kernel matrix which forms a basis for community partitioning. We further propose a novel robust multiple-kernel combination-based fuzzy clustering algorithm. Extensive experiments are conducted on many real-world networks that contain ground truth on community structures. The experimental results indicate that the proposed algorithm is more efficient than other existing community detection methods and related kernel clustering algorithms. This study demonstrates the feasibility and efficiency of the multiple-kernel learning method for community detection.