不完备目标信息系统中基于描述子的知识约简  

Knowledge Reductions Based on Descriptors in Incomplete Systems

作　　者：谢军[1,2] 杨习贝[1] 杨静宇[1] 孙怀江[1] 宋余庆[2] 

机构地区：[1]南京理工大学计算机科学与技术学院,南京210094 [2]江苏大学计算机科学与通信工程学院,镇江212013

出　　处：《计算机科学》2008年第10期181-185,共5页Computer Science

基　　金：国家自然科学基金(No.60472060,60572034);国家自然科学基金重点项目(No.60632050);江苏省自然科学基金(No.BK2006081)资助项目

摘　　要：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,根据描述子的定义和基于描述子的粗糙集模型,讨论了知识约简问题。给出了求得描述子所有约简的具体操作方法。根据描述子的支持集与决策类之间的关系,提出了描述子的下、上近似相对约简概念,并给出这两种约简的判定定理及区分函数,为从不完备信息系统中获取简化的决策规则提供了新的理论基础与操作手段。The incomplete information system in which all unknown values were looked as lost was deeply investigated. In such information system, approaches to knowledge reductions were studied from the viewpoint of the descriptor and descriptor-based rough set model. The practical approach to compute all reducts of a descriptor was presented. Based on the relationships between the support set of a descriptor and decision classes, the concept of descriptor-based lower and upper approximate relative reduets were proposed, the judgment theorems, discernibility functions with respect to these reducts were obtained, from which we obtained new theory and practical approach to derive simplified decision rules from the incomplete objective information system.

关 键 词：不完备目标信息系统 描述子 粗糙集 约简 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP37[自动化与计算机技术—控制科学与工程]