不完备不一致决策系统的最大分布约简及计算方法  

Maximum Distribution Reduct and Its Calculation Method in Incomplete Inconsistent Decision Systems

作　　者：蒙祖强[1] 许珂[1] 周石泉[1] 

机构地区：[1]广西大学计算机与电子信息学院,广西南宁530004

出　　处：《广西师范大学学报（自然科学版）》2011年第3期89-93,共5页Journal of Guangxi Normal University:Natural Science Edition

基　　金：国家自然科学基金资助项目(61063032);广西教育厅科研基金资助项目(201012MS010)

摘　　要：不完备不一致决策系统中,条件属性下的相容划分与决策属性下的等价划分形成了复杂的交集,导致出现了多种不同的约简概念,从而使约简问题变得更加复杂。本文将最大分布约简的概念引入不完备不一致决策系统中,然后研究其在不完备不一致决策系统中的性质,发现其核属性不具备传统约简核属性通常所具备的继承特性,因而不能通过增加属性的方法来计算此类约简。但是通过不断的属性测试和删除操作,成功地构造了不完备不一致决策系统中计算最大分布约简的算法,并给出了算法的描述和复杂度分析。通过实例分析,本文算法是有效的且具有实际意义。In inconsistent incomplete decision systems（IIDSs）,some tolerance classes in tolerance partition overlap more than one decision class,so as to produce complex overlapping subsets between tolerance partition and decision partition.This leads to occurrence of many concepts of reductions in IIDSs and makes the reduction problem more complex.Therefore,the concept of maximum distribution reduct is extended to IIDSs in this paper,then,some of its properties in IIDSs are analyzed.It is found that,unlike other reducts＇ core attributes,the maximum distribution reduct＇s core attribute does not have inheritable trait.This shows that the maximum distribution reduct can not be generated by adding attributes to core attribute set.But,by using the testing and deleting operations repeatedly,an algorithm for computing the maximum distribution reduct in IIDSs is successfully constructed in this paper.The algorithm＇s description and its complexity analysis are also given.Finally,the proposed algorithm is illustrated to be effective and be of practical significance through sample analysis.

关 键 词：决策系统 不一致性 不完备性 最大分布约简 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]