Research And Application Of Spectral Clustering

Times of Big data has come to us.In today's society,if you control data,you win.How to efficiently and accurately analyze and deal with massive data has become a hot topic nowadays.As a rising star in the field of data mining,clustering analysis has attracted a lot of attention in recent years and has achieved rapid development.Spectral clustering algorithm is a new kind of excellent algorithm based on the traditional clustering algorithm,combined with the spectral theory of graphs.It can be used to cluster the sample data of any shape and converge to the global optimal solution.This paper from the background,development status,the basic idea,specific processes and other problems are introduced in detail and based on the problems,we propose some new solutions.On the problem of how to select the feature vectors,the space formed by eigenvector determines the effect of spectral decomposition,and then determines the quality of clustering,so it is very important to select the appropriate feature vectors.In this paper,we propose a spectral clustering algorithm based on feature vector automatic selection,and the results of numerical experiments show that the proposed method can improve the quality of the algorithm and solve the problem of selecting the feature vector in the processing of the typical spectral clustering problem.On the problem of optimization algorithm,low rank subspace clustering algorithm is suitable for the clustering of defect data,but it may be ill-posed.In this paper,a regularized low rank subspace spectral clustering algorithm is proposed to solve the ill posed problem by regularization method and the numerical experiments show that the algorithm can solve the problem of spectral clustering in a certain extent,suppress noise and get high quality clustering results.