Spectral Clustering Algorithm Based On Nearest Neighbor Relation And Ensemble Clustering

Spectral clustering algorithms are widely used in computer vision,speech recognition,VLSI design,text mining,and other fields.Ensemble clustering can handle data that is difficult to handle by a single clustering algorithm,such as complex structures,fuzzy boundaries,and high-dimensional data.This paper proposes a spectral clustering algorithm and a weighted ensemble clustering algorithm based on the nearest neighbor relationship of data and the internal evaluation index of clustering.The specific research content is as follows:Firstly,a spectral clustering algorithm based on data proximity is proposed,which mainly studies the proximity of data,constructs a K-nearest neighbor graph.Then constructs ε Nearest neighbor graph based on the K-nearest neighbor graph to avoid disadvantage due to fixed K values,which is more suitable for show data distribution.In addition,it is proposed to use the nearest neighbor relationship of the data to set the distance weight,and use the mean and standard deviation of the distance between the data and its K-nearest neighbor to improve the kernel parameters A large number of data experiments have verified the efficiency of the proposed algorithm.Secondly,a weighted ensemble clustering algorithm is proposed,which mainly uses the cluster internal evaluation indicators to set weights for the base clusters,and adds thresholds to the construction of the co-occurrence matrix.A pair of data with fewer co-occurrence times is filtered,and a co-occurrence matrix is constructed to obtain clustering results using graph segmentation.Numerical experiments demonstrate the good performance of the algorithm.