基于类内类间离散度的分类器设计方法  

Classifier design based on between-class and within-class scatter

作　　者：郭亚琴[1] 王正群[2] 

机构地区：[1]紫琅职业技术学院,南通226002 [2]扬州大学信息工程学院,扬州225009

出　　处：《信息技术》2010年第5期35-37,共3页Information Technology

基　　金：国家自然科学基金项目资助(60875004);江苏省高校自然科学基础研究项目资助(07KJB520133);紫琅职业技术学院科研项目资助(科研2008003)

摘　　要：提出了一种基于类内类间离散度的最小距离分类器设计方法。该方法解决了分类过程中样本点分散和样本不可分问题。基本思想是:利用训练样本定义类内类间离散度矩阵,根据对离散度矩阵的分析,建立目标函数,求解目标函数,得到一组最优解。在分类时,应用最优解进行加权定义,从而获得更好的识别结果。采用UCI标准数据集实验,实验结果表明该算法具有一定的优越性。Minimum distance classifier based on between-class and within-class scatter is proposed.The new method can resolve sample dispersion and the unclassifiable region problems.The basic idea is to resolve an optimal resolution.First,between-class and within-class scatter matrix is defined by training sample.Then,objective function is set up.The optimal weight is obtained through solving objective function.By adding weight defining for distance in the classification phase,the classifier improved its classification accuracy.Experiment is tested on UCI standard database.The experimental results show that the proposed minimum distance classifier based on between-class and within-class scatter is effective.

关 键 词：分类 最小距离分类器 类内类间离散度 最优解 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]