基于差别信息树的roughset属性约简算法控制与决策.PDFVIP

第30 卷 第8 期 控 制 与 决 策 2015 年 8 月 Vol. 30 No. 8 Control and Decision Aug. 2015 文章编号: 1001-0920 (2015) 08-1531-06 DOI: 10.13195/j.kzyjc.2014.0724 基于差别信息树的rough set 属性约简算法 蒋 瑜 (成都信息工程大学软件工程学院，成都610225) 摘 要: 差别矩阵为粗糙集属性约简提供了很好的思路, 但差别矩阵中存在冗余的重复和父集元素. 为了消除这些 冗余元素, 提出一棵有序树: 差别信息树, 该树能消除差别矩阵中的重复元素, 同时在大多数情况下也能完全消除父 集元素, 实现对差别矩阵中非空元素的压缩存储. 为了验证差别信息树的有效性, 提出一种属性约简完备算法, 并使 2 该算法的时间复杂度降为(∣ ∣∣ ∣ ). 关键词: 粗糙集；差别矩阵；属性约简；差别信息树 中图分类号: TP18 文献标志码: A Attribute reduction with rough set based on discernibility information tree JIANG Yu (College of Software Engineering，Chengdu University of Information Technology，Chengdu 610225，China. E-mail: jiangyu@cuit.edu.cn) Abstract: Attribute reduction plays an essential role in rough set. Discernibility matrix provides a good way for attribute reduction in rough set, but there are many redundancy and pointless nonempty elements in discernibility matrix, such as duplications and supersets. In order to eliminate these redundancies and pointless elements, a discernibility information tree is proposed, which is an extended order-tree, and can be easy to fully and partly eliminate duplicates of elements and supersets of elements in discernibility matrix efficiently. In order to efficiently utilize the dicernibility information tree structure for 2 attribute reduction, an attribute reduction complete algorithm is developed, and it’s complexity is reduced to (∣ ∣∣ ∣ ). Keywords: rough set ；discernibility matrix ；attribute reduction ；discernibility