Community Detection of Complex Networks Based on the Spectrum Optimization Algorithm

This paper presents a spectrum optimization algorithm for community detection of the complex networks. An edge cutting model is proposed for selecting edges to be removed from candidates by minimizing algebraic connectivity function. In this model a greedy heuristic method is used to get the lower bound of optimal value, which makes it applicable to large-scale networks. Additionally, by weighting every edge based on Fielder vector, this model can effectively reduce the disadvantage influence of periphery edges of a graph on the result. The experiments on the simulated and the real complex networks show that this algorithm can reduce the time complexity of Newman-Girvan algorithm while preserving the performance of community detection.