约简加速求解的属性簇方法  

Accelerator for finding reduct based on attribute group

作　　者：陈妍 宋晶晶[1,2] 杨习贝 Chen Yan;Song Jingjing;Yang Xibei(School of Computer,Jiangsu University of Science and Technology,Zhenjiang 212003,China;Key Laboratory of Data Science and Intelligent Application of Fujian Province University,Zhangzhou 363000,China)

机构地区：[1]江苏科技大学计算机学院,江苏镇江212003 [2]数据科学与智能应用福建省高校重点实验室,福建漳州363000

出　　处：《南京理工大学学报》2020年第2期216-223,共8页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(61906078,61572242);数据科学与智能应用福建省高校重点实验室开放课题(D1901)。

摘　　要：为了进一步提高约简求解的效率,该文在桶模型的基础上,从数据中属性间的相似性程度出发,将属性划分为不同的簇,使得在约简的搜索进程中,只需以属性簇为基准进行候选属性的筛选即可达到压缩属性搜索空间的目的。实验结果表明,无论是采用传统的邻域计算或是基于桶模型的邻域计算,在不降低分类性能的前提下,基于属性簇的搜索策略都能显著降低求解约简的时间消耗。该文研究可从样本和属性两方面为约简求解加速提供参考。To improve the time efficiency of obtaining the reducts,based on the mechanism of bucket model,attributes are divided into different groups by considering the similarity between attributes.It follows that in the searching process of deriving reducts,attributes out of those groups containing at least one attribute in potential reducts should be evaluated,which can effectively reduce the searching space.Compared with the forward greedy searching approach,the experimental results show that the proposed strategy can significantly reduce the time consumption of obtaining reducts and the classification performance of reducts derived by using the strategy of attribute group is not decreased.This study provides a useful idea for accelerating the process of finding reducts.

关 键 词：属性簇 属性约简 桶模型 邻域粗糙集 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]