扩散先验分布下Bayes多总体分类判别方法的构造  

The design of bayesian classification and discrimination method among several populations based on diffuse prior distribution

作　　者：朱慧明[1] 韩玉启[1] 

机构地区：[1]南京理工大学管理学院,江苏南京210094

出　　处：《数理统计与管理》2003年第1期33-37,共5页Journal of Applied Statistics and Management

摘　　要：本文研究了各总体服从多元正态分布 ,其未知参数的先验分布均为扩散先验分布时 ,如何利用待判样品的预报密度函数、构造后验概率比并据此对样品进行分类与判别 ;此方法并不需要假设各总体分布的协方差相同 ,而且在预试样本容量较小时仍然可行。In this paper,the authors study a method on how to classify a sample into one of the several known normal populations in terms of posterior probability ratio established by the samples predictive density functions when the unknown parameters prior distributions are diffuse prior distribution.The method doesnt require the consistency of each populations covariance and it remains valid when the training sample size is not very large.

关 键 词：扩散先验分布 Bayes判别分析 多元统计 样品 后验概率 分类判别规则 

分 类 号：O212.8[理学—概率论与数理统计]