SAWEG: A new algorithm for network clustering

Abstract

The relationship between nodes in the network is reflected by edges. It is novel to study the network from the perspective of edges rather than nodes. Following the idea of edge graph clustering which provides an unorthodox approach to represent the topology of the systems, we propose a new clustering method called Spectral Analysis based on Weighted edge graphs to find the overlapping clusters. According to the incidence matrix of the original network, we obtain the corresponding edge graph. The first two nontrivial eigenvectors dimensional spaces are built with the Laplacian Matrix, and we get edge dendrogram of the network. Then cut the edge dendrogram by using the improved partition density to get the optimal community structure. The proposed algorithm successfully finds the common nodes between clusters. Experiments on five real-world networks show that the proposed SAWEG algorithm performs better than the other three benchmarking clustering algorithms.