Spectral Based Eigenvector Selection Method for Community Detection

Abstract

Detecting communities in complex networks is an important problem. Spectral method has been successfully used to solve data clustering and community detection problems. However getting the proper eigenvectors is usually di‐cult and depends on extra thresholds. In this paper, we devise an efiective method to choose suitable eigenvectors. The method needs no extra thresholds and can get the reasonable communities in the network. Furthermore we can identify the overlapping communities in complex networks by clustering the chosen eigenvectors with fuzzy c-means algorithm. Experimental results show that the algorithm is efiective in detecting communities.