摘要
针对变量比较少的判别分析问题,讨论了一种基于变量择优的Fisher逐步判别分析方法.在Fisher判别分析方法和变量空间的多项式扩展方法的基础上,该方法结合文中定义的判别效率函数,给出了挑选变量的准则.为了检验新方法的效果,分别通过一个实例、5个UCI数据集和3个模拟数据集进行测试,并与Fisher判别分析法和Fisher逐步判别分析法的效果作对比.结果表明,对变量比较少的判别问题,文中方法有助于提高判别模型的正确率.
In this paper,a Fisher stepwise discriminant analysis method based on selecting optimal variables is discussed for the discriminant problem when the number of variables is small.On the basis of the Fisher discriminant analysis method and the polynomial expansion method of variable space,this method combines with the discriminant efficiency function defined in the paper,and gives the criteria for selecting variables.To examine the effect of the new method,one example,five UCI data sets and three simulation data sets are tested,and the effects of this method are compared with those of Fisher discriminant method and Fisher stepwise discriminant analysis method.The results show that this method is helpful to improve the correct ratio of the discriminant model when the number of variables is small.
作者
黄利文
HUANG Liwen(College of Mathematics and Computer Science,Quanzhou Normal University,Quanzhou 362000;Key Laboratory of Financial Mathematics(Putian University),Fujian Province University,Putian 351100)
出处
《系统科学与数学》
CSCD
北大核心
2021年第8期2338-2348,共11页
Journal of Systems Science and Mathematical Sciences
基金
国家自然科学基金(52070119)
金融数学福建省高校重点实验室(莆田学院)开放课题(JR201801)资助。
关键词
判别分析
判别模型
逐步判别分析
变量择优
Discriminant analysis
discriminant model
stepwise discriminant analysis
selecting optimal variables
