Hash快速属性约简算法  

Quick Attribute Reduction Algorithm with Hash

作　　者：刘勇[1,2] 熊蓉[1,2] 褚健[1,2] 

机构地区：[1]浙江大学工业控制国家重点实验室,杭州310027 [2]浙江大学智能系统与控制研究所,杭州310027

出　　处：《计算机学报》2009年第8期1493-1499,共7页Chinese Journal of Computers

基　　金：国家自然科学基金(60803053;60675049);浙江省自然科学基金(Y106414);国家"八六三"高技术研究发展计划项目基金(2008AA04Z209);国家博士后科学基金(20081459);武器装备预研基金(9140A06050609JW0402)资助~~

摘　　要：从决策系统的不一致情况出发,给出了不一致度的概念及其性质,并证明了不一致记录与正区域的等价关系.在此基础上,提出了基于Hash的正区域计算方法,时间复杂度下降为O(|U|);利用不一致情况的性质设计了一个基于不一致记录数的属性重要性测量参数,用新的测量参数设计了一个基于二次Hash的约简算法,其复杂度下降为O(|C|2|U/C|),并证明采用该测量参数所获得约简的完备性.最后通过实验证明该文正区域算法和约简算法的高效性.This paper presents the concept and property of inconsistency from the inconsistent condition of decision system. It also presents the relationship between the positive region and in- consistent records. A hash based algorithm calculating positive region has been presented and its temporal complexity decreases to O（ ｜U｜ ）. Based on the characteristics of inconsistency, a new attribute measure has been introduced, then a corresponding reduction algorithm with twice-hash is presented, and its temporal complexity is O（ ｜ C｜^2 ｜ U/C｜ ）, this paper also proves this algorithm is complete. The efficiency of the algorithms is proved by the experiments.

关 键 词：粗糙集 正区域 约简 HASH 不一致度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]