基于相对辨识关系的属性约简算法  

An Attribute Reduction Algorithm Based on Relative Discernible Relation

作　　者：孙滨 毛建景 SUN Bin;MAO Jian-jing(School of Information Technology,Zhengzhou University of Industrial Technology,Zhengzhou 451100,China)

机构地区：[1]郑州工业应用技术学院信息工程学院,河南郑州451100

出　　处：《计算机技术与发展》2018年第4期99-103,共5页Computer Technology and Development

基　　金：河南省重大科技专项(162102210319);河南省基础与前沿技术研究计划项目(142300410283);河南省科学技术研究重点项目(12B520063;14B520065);河南省高等学校青年骨干教师资助计划项目(2016GGJS-182)

摘　　要：介绍了决策信息系统中的可辨识关系、相对辨识关系等相关概念,完成了属性集独立性判定工作。将属性集的可辨识能力和相对辨识能力与属性集所辨识的对象联系起来,研究了两种基于相对辨识关系的属性重要度求解方法,完成了属性集独立或依赖、是否为决策信息系统约简的判定。依据相对辨识关系,给出了相应的改进算法,利用该算法描述因属性集属性的增减而引起它的相对辨识能力的变化。该算法是从条件属性集中先判断各个属性的相对可辨识关系,其对象对个数最大者组成约简集,然后再在此约简集中逐渐添加属性,直至满足约简的条件。该算法是一种无核的属性约简算法,无论算法时间复杂度还是约简工作量在一定程度上都有所降低,并通过实例验证了其有效性。We mainly introduce some related concepts in decision information system,such as discernible relation and relative discernible relation,and complete the attribute set independence judgment.Connecting the discernibility and relative discernibility of the attribute set with the object of identification of the property set,we also study the evaluation methods of attribute importance between the two kinds of relative identification and finish the judgment of attribute's independence or dependence and whether to do reduction of decision information system.Based on the relative recognition relation,the corresponding improvement algorithm is given,by which the variation of the relative recognition caused by the addition or subtraction of attribute set attributes is described.The algorithm determines the relatively discernible relation of each attribute from the condition attribute set,taking the largest number of object as the reduction set where the attributes are gradually added until meeting the conditions of reduction.It is a kind of seedless attribute reduction algorithm,which is reduced to a certain extent regardless of time complexity or reduction workload.Its validity is verified by an example.

关 键 词：粗糙集 相对可辨识关系 属性集 属性约简 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]