Community Detection Based On Optimization Of Spectral Clustering

As one of the core contents in complex network research,there has been multiple sophisticated solutions in community detection.Known as a typical clustering method,spectral clustering is widely used in community detection with its distinct data-mapping logic and superior multidimensional data clustering ability.There're deficiencies in traditional spectral community detection.On the one side traditional spectral detection methods replace similarity analysis with node connectivity,which results in potential precision shortage.On the other side,pre-known node's adscription information often exists during community detection,which could assist spectral clustering and enhance detection efficiency.This paper present a semi-supervised spectral clustering optimization method based on signal transmission according to existing similarity deficiencies in spectral clustering community detection.This method constructs node similarity with signal transmission,meanwhile introduces pre-known knowledge into clustering according to semi supervised learning theory.Experiments proves the method could improve detection by fully propagates signal in network and build similarity matrix with node's signal value as the optimization of spectral clustering.This paper present a spectral clustering optimization method based on signal diffusion according to the optimization strategy on similarity precision and time efficiency in common community detection.The method solves the defect of unfixed signal amount in signal transmission,and proposes a simplified similarity construction to improve the performance of spectral clustering due to its deficiency.Experiments proves this algorithm significantly reduces the cost of Laplacian construction,and reach better community detection efficiency.