关键词: 密度敏感,相似矩阵,秩约束,聚类

Abstract: The traditional clustering algorithms generally use Euclidean distance to acquire the similar matrix.In some more complex data processing,Euclidean distance doesn’t have the ability of describing the characters of data because it can’t reflect the global consistency.An adaptive clustering algorithm based on rank constraint density sensitive distance (RCDSD) was proposed in this paper.First,a density sensitive distance similarity measure is introduced to acquire the similar matrix which enlarges the distance between the different classes and reduces the distance between the same classes effectively,so as to solve the disadvantages of clustering results deviation of the traditional clustering algorithm based on Euclidean distance.Second,the rank constraint is imposed to the Laplacian matrix of the similarity matrix,thus the number of connected area of the similar matrix is equal to the number of clustering,and the data can be directly divided into the right class and the algorithm can take the final clustering result,while the algorithm does not need to perform k-means or other discrete procedure.Experimental results show that the approach can obtain accurate clustering results and improve the clustering performance on both artificial simulation data sets and real data sets.

Key words: Density sensitive,Similarity matrix,Rank constraints,Clustering