基于混合高斯模型的指标优化方法  

The Method of Index Optimization Based on Gaussian Mixture Model

作　　者：陈超[1] 姜鑫[1] 沙基昌[1] 

机构地区：[1]国防科学技术大学信息系统工程重点实验室,湖南长沙410073

出　　处：《数学的实践与认识》2011年第15期151-155,共5页Mathematics in Practice and Theory

基　　金：教育部博士点基金(20104307120020);国家自然科学基金(71071160)

摘　　要：借鉴混合高斯模型的方法,提出了基于混合高斯模型的非参数检验法用于评价指标体系的构建.首先运用混合高斯模型对现有指标体系进行聚类分析,然后运用非参数检验的秩相关系数对分类后的指标体系进行筛选,最终确定评价指标体系的构成,保证指标体系的全面性和代表性.实例结果表明,方法分类精度高、简便、实用,在指标优化中具有应用价值.This article proposed a non-parameter test method by using Gaussian Mixture Model idea for reference to evaluate the construction of index architecture. First Gaassian Mixture Model is used to cluster analysis the index architecture, second rank correlation coefficient of non-parameter test method is used to select index architecture, and then the evaluate index architecture is determined to guarantee the comprehensiveness and representativenese of the index architecture. The example shows that the new method is simple; practical and having high classification precision, and it is also useful for index optimization.

关 键 词：混合高斯模型 非参数检验 指标优化 聚类 

分 类 号：O212.1[理学—概率论与数理统计]