多元线性回归模型的聚类分析方法研究

Classification for Multiple Linear Regression Methods

提出一种对大量的多元线性回归模型进行聚类分析的方法。首先利用增广矩阵的相关系数矩阵定义了2个多元回归模型之间的距离以及模型集合的质心和半径等相关概念。然后采用Squeezer聚类方法,以过程全自动化的方式,实现对多元线性回归模型集合进行聚类分析。通过仿真研究验证了方法的有效性,取得满意的分析结果。

A cluster analysis method on massive multiple linear regression models was proposed. Firstly, the concepts such as distance between two multiple regression models, the centroid and radius of multiple linear regression model set were defined by using the correlation coefficient matrix of augmented matrix. Then Squeezer cluster method was applied to realize the cluster analysis on multivariate linear regression models based on entire automation of the process. The simulation case confirms validity of this method and lead to satisfactory results.