基于相关阵的聚类分析  

A cluster method based on correlation matrices

作　　者：沈娟华[1] 汪四水[1] 

机构地区：[1]辽宁科技大学理学院,辽宁鞍山114051

出　　处：《辽宁科技大学学报》2008年第6期601-605,610,共6页Journal of University of Science and Technology Liaoning

摘　　要：聚类分析是一类很重要的统计分析方法,有着广泛的应用。利用相关矩阵来进行聚类分析,对于高维的数据,将数据表示成相关矩阵,然后通过相关矩阵的函数变换来进行聚类。通过对一类相关阵迭代的Concor算法的探讨,给出了该方法的相关性质及利用该算法的聚类方法,并结合Matlab编程,给出了三、四、五阶Concor算法矩阵序列的收敛不动点。Cluster analysis is an important method which is widely used. Correlation matrices were used to make cluster analysis. The relation of multidimensional data was described by the correlation matrices, and the data were clustered through transforming correlation matrices. A method named Concor was discussed and carried out by iterating correlation matrices. Some relative properties of the method are presented, and the Matlab program is introduced. Fixed matrices based on Concor method are put forward when the rank is three, four and five.

关 键 词：Concor聚类 变换 完美对称矩阵 

分 类 号：O212[理学—概率论与数理统计]