期刊论文

中文

模糊系统与数学

1001-7402

2007

21

1

128-136

10.3969/j.issn.1001-7402.2007.01.021

Supported by National Natural Science Foundation of China （60474023） Scientific Research Fund of Hunan Provincial Edducation Pepartment（06C097） Science and Technology Key Project Fund of Ministry of Education（03184） Major State Basic Research Development Program of China（2002CB312200） China Postdoctoral Science Foundation（2005037316） Research Start Foundation of Changsha University of Science and Technology（52129905）

本校为第一机构

给出一种新的属性依赖性度量,根据这种依赖性度量,给出属性重要性的定义,然后以这种属性重要性为权重,给出一种基于加权综合的样本分类算法.由于属性的重要性是由条件属性相对于决策属性的依赖性决定的,它的数值相对比较客观,这样解决了常用的定权方法的弱点,它们一般是凭经验或由专家给出,具有相当的主观性.

In this paper, first, we give a new measure of attribute dependency. According to this measure, we give a definition of significance of attribute. Then a comprehensive weighted classification algorithm is put forward. Because the significance of attribute is computed on the basis of dependency between condition and decision attributes, its value is objective. This method compensate the drawback th...