不完备系统中基于优势关系的最优可信规则获取

Mining Optimal Credible Rules in Incomplete Systems Based on Dominance Relation

以具有遗漏型未知属性值的不完备目标信息系统为研究对象,根据优势关系,提出升序和降序最优可信规则.给出基于优势关系的描述子的概念,由此获得不完备目标信息系统中的升序和降序可信规则.定义了基于区分矩阵的区分函数用以计算对象的升序和降序约简,从而获得升序和降序最优可信规则,为从不完备目标信息系统中获取规则提供了新的理论基础与技术手段.

In this paper, the incomplete objective information system in which all unknown values are considered as lost is deeply investigated. In such information systems, the concepts of optimal credible rules in terms of increasing and decreasing preferences are proposed based on the dominance relation. To derive all credible rules from the incomplete objective information system, the concept of descriptions are presented. Furthermore, from the viewpoint of the discernibility functions based on the discernibility matrixes, the increasing and decreasing reducts of objects can he worked out and as a result, the optimal credible rules are obtained. The research is meaningful both in the theory and in applications for the acquisition of rules in the incomplete objective information systems.