基于哈希桶的快速三支决策邻域分类器  

Fast Three-way Decision Neighborhood Classifier Based on Hash Bucket

作　　者：贾润亮[1] 张海玉[1] JIA Runliang;ZHANG Haiyu(School of Finance and Economics,Taiyuan University of Technology,Taiyuan 030024,China)

机构地区：[1]太原理工大学财经学院,太原030024

出　　处：《小型微型计算机系统》2025年第4期776-782,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61403271)资助.

摘　　要：三支决策邻域分类器作为邻域粗糙集的重要扩展,目前已成为数据挖掘中一种有效的分类方法.然而,三支决策邻域分类器当前仍存在两方面的局限,一是获得测试样本邻域类的计算复杂度较高,二是测试样本对于多个最大决策类场景无法确定最终的类别标签,为了解决此问题,本文提出一种基于哈希桶方法的快速三支决策邻域分类器.首先,对分类训练集通过哈希规则将样本对象映射到对应的哈希桶中,通过哈希桶实现了邻域的搜索范围被限制在对象所属桶和相邻两个桶中;然后,为了避免测试样本针对多个最大决策类存在类别无法判定的情况,定义一种平均距离度来描述对象与决策类之间的距离程度,在多数投票规则基础上结合平均距离度,实现了测试对象对最大决策类的识别能力;最后,综合快速邻域类计算和平均距离度,建立了基于哈希桶的快速三支决策邻域分类器模型.实验结果表明了所提出的分类器具有较好的分类性能和分类效率.As an important extension of neighborhood rough sets,the three-way decision neighborhood classifier has become an effective classification method in data mining.However,the three-way decision neighborhood classifiers currently have two limitations:firstly,the computational complexity of obtaining neighborhood for test samples is high,and secondly,the test samples cannot determine the final class labels for multiple maximum decision class scenarios.To address this issue,a fast three-way decision neighborhood classifier based on hash bucket method is proposed in this paper.Firstly,for the classification training set,the sample objects are mapped to the corresponding hash buckets through hash rules,and the search range of the neighborhood is limited to the bucket to which the object belongs and the adjacent two buckets through hash buckets;Then,in order to avoid the situation where the test sample cannot determine the category for multiple maximum decision classes,an average distance measurement is defined to describe the distance between the object and the decision class.Based on the majority voting rules,the average distance measurement is combined to achieve the recognition ability of the test object for the maximum decision class.Finally,a fast three-way decision neighborhood classifier model based on hash buckets was established by combining fast neighborhood calculation and average distance measurement.The experimental results indicate that the proposed classifier has good classification performance and efficiency.

关 键 词：邻域粗糙集 邻域分类器 哈希桶 三支决策 平均距离度 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]