一种基于序贯三支决策的属性约简方法研究  

An Attribute Reduction Method Based on Sequential Three-way Decision

作　　者：盛茹雪 李红宇[1] 姜春茂[1] 郭豆豆 SHENG Ru-xue;Li Hong-yu;JIANG Chun-mao;GUO Dou-dou(School of Computer Science Technology and Information Engineering,Harbin Normal University,Harbin 150025,China)

机构地区：[1]哈尔滨师范大学计算机科学与信息工程学院,黑龙江哈尔滨150025

出　　处：《模糊系统与数学》2021年第6期48-65,共18页Fuzzy Systems and Mathematics

基　　金：黑龙江省自然科学基金资助项目(LH2020F031);哈尔滨师范大学研究生创新项目(HSDSSCX2020-30)。

摘　　要：属性约简一直是粗糙集理论研究的热点问题。近年来,基于启发式的属性约简方法研究逐步兴起,其核心内容是在约简过程中增添最重要的属性,但此类算法存在复杂度高和时间消耗大等问题。针对此,提出一种基于序贯三支决策的属性约简方法。将属性作为划分策略的治略对象,依据属性重要度进行三支治略:根据属性重要度和阈值将属性分为正划分集合、负划分集合和延迟划分集合。发生正划分的属性集合被接受添加到决策属性集,负划分的属性集合被拒绝添加到决策属性集,对延迟决策的属性集合重复以上操作直至约简结果满足约束条件。此算法根据划分结果选取加入决策属性集的元素集合,从而大大降低了时间消耗。实验选取了8组UCI数据集,在传统全局约简条件和集成约简条件下分别进行,实验结果表明,本文提出算法在两种条件下能够在保证分类精度的前提下,有效降低时间消耗。Attribute reduction is a hot issue in rough set theory.The research of attribute reduction method based on heuristic is gradually rising in recent years.Its core content is to add essential attributes in the reduction process,but such algorithms have high complexity and time consumption.Given this,this paper proposes an attribute reduction method based on sequential three-way decision.Attributes are used as the object of acting enforcement in the sequential three-way decision model,and the three-way strategy is performed according to the significance of the attribute:According to attribute significance and threshold,attributes are divided into positive divide set,negative divide set and delayed divide set.The attribute set with positive divide is accepted to be added to the decision attribute set,and the attribute set with negative divide is rejected to be added to the decision attribute set.Repeat the above operation for the attribute set with delayed decision until the reduction result meets the constraint conditions.This algorithm selects the elements sets added to the decision attribute set according to the result of the divide,thus significantly reducing the time consumption.The experiment selected eight sets of UCI data sets,carried out under the traditional global reduction conditions and the local reduction conditions.The experimental results show that the algorithm proposed in this paper can effectively reduce the time consumption under the premise of ensuring the classification accuracy under the two conditions.

关 键 词：三支决策 序贯三支决策 邻域粗糙集 属性约简 属性重要度 

分 类 号：O159[理学—数学]