一致完备决策系统中属性重要性定义的改进  

A New Definition of Attribute Significance in Consistent Complete Decision Systems

作　　者：连钢[1] 程媛媛 吴根秀[1] 

机构地区：[1]江西师范大学数学与信息科学学院,江西南昌330022 [2]江西省上饶市第一中学,江西上饶334000

出　　处：《江西师范大学学报（自然科学版）》2008年第3期326-329,共4页Journal of Jiangxi Normal University(Natural Science Edition)

基　　金：江西省自然科学基金(2007GZS0186);江西省教育厅科技(2008GJJ0815);江西省研究生创新专项资金(07A045);江西师范大学研究生创新基金资助项目

摘　　要：在一致完备决策表中利用代数(信息熵)定义比较属性之间的重要性时往往会得到一些与实际不相吻合的结论,这种问题产生的主要根源是其定义本身的不完善所造成的.该文从产生问题的根源出发,从约简的角度重新定义了决策表中的各个属性(属性子集)的重要性,同时还给出了体现2个不同属性之间相互关系的相关系数的定义.最后通过对2个不同的例子来说明新定义的重要性比代数(信息熵)定义更具有合理性.Some conclusions that are inconsistent with the practice may appear when making comparison between attributes among complete consistent decision tables. The main source of this problem is the imperfections caused by its own definition. In this paper,after analyzing the defect of original definition and in the point of reduction view,a new definition of significance of each attribute and subset of attributes in decision tables are given. What＇ s more, the definition of correlation coefficient of two different attributes is given. Finally, two examples illustrate that the significance defined in this paper is more reasonable than that of algebraic definition or definition of entropy.

关 键 词：属性约简 属性重要性 决策系统 相关系数 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]