摘要
基于粗糙集理论定义的属性约简大都要求约简前后正区域保持不变或者非负区域不变.在概率型决策粗糙集模型下,决策区域和决策规则与属性增减之间并不具备单调性.因此,决策者基于约简后的属性集合所作的决策风险最小就变得非常有意义.针对这种与各个区域无关的基于决策风险最小化的属性约简进行了研究.考虑到不同属性对决策表的决策分类能力不同,提出了基于决策粗糙集模型的属性重要性概念,设计了一种有效的基于属性重要性的决策风险最小化启发式属性约简算法.实例分析与对比实验结果说明新方法是有效的.
The most important part of the classical rough set theory is upper approximate set and lower approximate set which is defined by accurate set inclusion.However the data at present is not accurate.As we all known,the Pawlak algebra rough set model is too rigid to lack fault-tolerance capability.In order to solve this problem,a bunch of probabilistic rough set models were proposed.Among these models the decision rough set model simulated better than others in human intelligence solving problems from semantic perspective. Attribute reduction is one important research field in rough set theory.The decision region,decision rule and the increase or decrease of attributes are not following monotonicity in the models of probability decision rough set,the positive region,negative region and border region of decision table maybe change differently before and after attribute reduction,therefore it ＇ s a important problem to evaluate attribute reduction is appropriate according to regions change.Therefore the definition of attribute reduction based on risk minimization under decision rough set model was proposed in order to solve this problem. The definition of attribute reduction based on rough set theory mostly requires that positive or nonnegative region are the same as before.Decision region,decision rule and the increase or decrease of attributes are not following monotonicity in the model of probability decision rough set.Therefore,it is very meaningful to minimize the decision risk according to attributes sets after reduction for decision makers.This paper studies the attribute reduction based on the minimum risk decision which has nothing to do with every region.Considering different attributes has different abilities to decide and make classification to decision table,an attribute significance concept based on decision rough set model was proposed,then we propose an effective decision-making risk minimization heuristic attribute reduction algorithm based on attribute significance.Examples analysis and experiment results comparison show the new method is effective.
出处
《南京大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第2期210-216,共7页
Journal of Nanjing University（Natural Science）
基金
国家自然科学基金(61073146
61272060)
重庆市自然科学基金(cstc2011jjA40045)
关键词
属性约简
风险最小化
属性重要性
决策粗糙集
attribute reduction
minimum decision cost
significance of attributes
decision-theoretic rough set
