Research Of Spectral Clustering Algorithm Based On Density Sensitivity

With the digitization of society,there is an urgent need to be able to handle large amounts of data,the emergence of data mining has promoted the development of information technology.In data mining,use clustering algorithms to discover knowledge that is hidden in large amounts of data without learning information.Spectral Clustering algorithm is one of clustering algorithm,the problem of data clustering in algorithm is transformed into the segmentation problem of undirected graphs,the algorithm focuses on the interrelationship between data points,weakening the characteristics of the data points themselves,and solving the problem that traditional clustering algorithms can not deal with complex clusters and easily fall into local optimal solutions.As a kind of spectral clustering algorithm,a Density Sensitive Spectral Clustering algorithm is proposed to describe how to construct a description method to describe the relationship between data samples.We define a similarity method to describe the data points: density-sensitive similarity,More effectively describes the similarity between data points,improve the accuracy of the algorithm.Based on the research of Density Sensitive Spectral Clustering algorithm,the advantages and disadvantages of Density Sensitive Spectral Clustering algorithm are analyzed,and the defect problem is optimized.The main work of this thesis is as follows:1.The clustering algorithm and algorithm flow of Density Sensitive Spectral Clustering algorithm are studied..The advantages of density sensitive distance and the existing problems in the algorithm are studied by theoretical analysis and simulation experiment.The scaling factor of density sensitive distance needs to be specified manually.The similarity between the data points with relative relative positions can be changed due to the change of the distance.The number of clusters needs to be specified manually.2.For the problem of the scaling factor and the density sensitive distance in the Dnsity Sensitive Spectral Clustering algorithm,the reason why the relative position of the data point is changed and the similarity relation is changed is calculated by calculating the length of the adjustable line segment.The local information is used to calculate each The Dnsity Sensitive Spectral Clustering algorithm with local information is proposed,and the algorithm is validated and validated on the data set.The advantages of the algorithm are more effective than the density sensitive spectral clustering algorithm.3.The influence of the number of clusters on the algorithm is studied.The relationship between the eigenvalues and the number of clusters in the Laplace matrix is analyzed.The method of determining the number of clusters according to the characteristics of the intrinsic gap is studied and studied in the data set.