基于系统熵的属性约简的简化差别矩阵方法  

Method of discernibility matrix for attribute reduction based on system entropy

作　　者：王熊彬[1] 郑雪峰[1] 徐章艳[2] 

机构地区：[1]北京科技大学信息工程学院,北京100083 [2]广西师范大学计算机系,广西桂林541004

出　　处：《计算机应用研究》2009年第7期2460-2464,共5页Application Research of Computers

基　　金：国家自然科学基金重点项目(69835001);广西教育厅基金资助项目

摘　　要：基于系统熵的属性约简是一种新型的属性约简。该模型由于同时考虑了条件属性集和决策属性集对决策表的分类能力,它是一种考虑较周全的属性约简模型。为设计高效的属性约简算法,首先引入简化差别矩阵,同时给出了基于该简化差别矩阵的属性约简定义,并证明该定义与基于系统熵的属性约简定义等价;然后用简化差别矩阵设计了一个基于系统熵的完备属性约简算法;最后用实例说明了新算法。Attribute reduction based on system entropy is the new attribute reduction. This is a more thorough considered model because the classfication ablilities of condition attributes and decision attributes to the decision table are considered. To design an efficent algorithm of attribute reduction based on the system entropy, first proposed the simplified discernibility ma- trix. At the same time, gave the definition of attribute reduction based on the simplified discernibility matrix. And proved that this new definition of attribute reduction is equal to the definition of attribute reduction based on the system entropy. Then designed a complete algorithm of attribute reduction based on system entropy with the new simplified discernibility matrix. At last, used an example to illustrate the efficency of the new algorithm.

关 键 词：粗糙集 系统熵 简化差别矩阵 属性约简 完备算法 复杂度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]