摘要
在不完备信息系统中,提出可信决策规则以及最优可信决策规则的概念,用以刻划信息系统中必定存在或可能存在的确定性决策规则.给出关于容差关系的最大相容类的∨-描述的概念,并由此得到初等可信决策规则.针对协调最大相容类,定义一种基于区分矩阵的区分函数;通过计算协调最大相容类的区分函数的析取范式,可由初等可信决策规则求得所有最优可信决策规则.最后给出条件属性的E-相对约简的概念及其区分函数求法.E-相对约简是保持不完备信息系统的确定性信息的极小条件属性子集.
We propose the concepts of credible decision rule and optimal credible decision rule in incomplete information systems to describe the certainly or possibly existing definite decision rules. In order to derive all the optimal credible decision rules from incomplete information systems, we introduce the ∨-descriptions of maximal tolerance classes and get the initial credible decision rules, and define a kind of discernibility function based on the discernibility matrix for the consistent maximal tolerance classes. By computing the disjunction normal form of the discernibility function for each consistent maximal tolerance class, we can get all the optimal credible decision rules from the initial credible decision rules. At last, we propose the concept of E-relative reduct of condition attributes and give its discernibility function computing approach. The E-relative reduct is the minimal condition attributes subset, which keeps the definite information contained in the original information system.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2005年第12期76-82,共7页
Systems Engineering-Theory & Practice
关键词
不完备信息系统
粗糙集
最大相容类
属性约简
区分函数
incomplete information system
rough set
maximal tolerance class
attribute reduct
discernibility function