基于优势关系粗糙集的动态容错分级决策模型  

Sorting Decision Model for Dynamic Fault Tolerance Based on Dominance Relation Rough Set

作　　者：苟光磊[1,2,3] 王国胤[1,2,4] 李鸿[2,4] 

机构地区：[1]西南交通大学信息科学与技术学院,四川成都610031 [2]中国科学院重庆绿色智能技术研究院,重庆401122 [3]重庆理工大学计算机科学与工程学院,重庆400054 [4]重庆邮电大学计算智能重庆市重点实验室,重庆400065

出　　处：《西南交通大学学报》2014年第1期147-152,共6页Journal of Southwest Jiaotong University

基　　金：国家自然科学基金资助项目(61073146);重庆市自然科学基金资助项目(cstc2012jjA40032)

摘　　要：为提高优势关系粗糙集模型在分级决策问题中的容错能力,将容错处理视为可动态调整的过程,根据用户向上、向下和综合两者的3种偏好趋向,提出了3种对应的分级算法,对边界域对象进行初始分级,利用对象的覆盖信息作为启发式知识调整其分级决策的结果,实现正确分级或接近正确分级.与变一致性优势关系粗糙集模型相比,不需要事先根据经验确定和调整阈值.案例应用结果表明:本文提出的3种偏好情况下的分级正确率比现有的分级算法平均提高了21.34%,对应的误分总代价平均降低了50.91%.To enhance the fault-tolerant capacity of the dominance relation rough set model in solving sorting decision problems, three efficient sorting decision algorithms are proposed by regarding the fault-tolerant processing as a dynamic adjusting process according to the fault tolerance direction of the user＇s preference, i. e. , upward, downward, or synthesis of the both. The boundary objects are initially ranked by the proposed algorithms, and the obtained results are adjusted using the coverage information as the heuristic criteria to achieve a accurate or near accurate sorting of the object finally. In contrast to the variable-consistency dominance-based rough set approach （VC-DRSA）, the proposed algorithms do not need prior domain knowledge to determine and adjust a threshold. Application of the algorithms to wine quality dataset show that the proposed methods can achieve a 21.34% improvement in average sorting accuracy and a 50. 91% reduction in average mis-sorting cost, compared with the existing methods.

关 键 词：粗糙集 优势关系 容错 分级 决策 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]