一种序决策信息系统中的快速属性约简算法  

A fast attribute reduction algorithm for sequential decision information system

作　　者：张义宗 王磊[1,2] 徐阳 王诚彪 ZHANG Yizong;WANG Lei;XU Yang;WANG Chengbiao(School of Information Engineering,Nanchang Institute of Technology,Nanchang 330099,China;Jiangxi Province Key Laboratory of Water Information Cooperative Sensing and Intelligent Processing,Nanchang 330099,China)

机构地区：[1]南昌工程学院信息工程学院,南昌330099 [2]江西省水信息协同感知与智能处理重点实验室,南昌330099

出　　处：《成都理工大学学报（自然科学版）》2023年第6期767-774,781,共9页Journal of Chengdu University of Technology: Science & Technology Edition

基　　金：国家自然科学基金地区项目(61562061);江西省教育厅科技项目(GJJ2202005,GJJ219110)。

摘　　要：为了提高无核或少核序决策信息系统中现有属性约简算法的执行效率,本文以知识粒度表征的属性重要度为启发信息并结合前向属性约简方法提出了一种新的属性约简算法。首先,介绍优势粗糙集方法的相关基础知识,并将经典粗糙集中基于知识粒度的属性约简算法引入优势粗糙集方法中,得到可处理序决策信息系统的属性约简算法;然后,通过分析序决策信息系统中知识粒在属性数目变化条件下的粗化与细化过程,得出相对冗余属性的判断定理,由此结合前向属性约简方法设计了快速属性约简算法;最后分析比较了2种算法的时间复杂度并选取了6个不同的UCI数据集进行算法性能的测试,测试结果表明,本文提出的算法比现有的属性约简算法高效。Based on the attribute importance of knowledge granularity representation and the forward attribute reduction method,a new attribute reduction algorithm is proposed in order to improve the efficiency of existing attribute reduction algorithms in non-kernel or less kernel order decision information systems.Firstly,the basic knowledge of the dominant rough set method is presented and the attribute reduction algorithm based on knowledge granularity of classical rough set is introduced into the dominant rough set method,thus the attribute reduction algorithm of processable decision information system is obtained.Then,by analyzing the coarsening and thinning process of knowledge granules in the order decision information system under the condition of variable attribute number,the judgment theorem of relative redundant attribute is obtained,and the fast attribute reduction algorithm is designed by combining forward attribute reduction method.Finally,the time complexity of the two algorithms is analyzed and compared,and 6 different UCI data sets are selected to test the algorithm performance.The test results show that the proposed algorithm is more efficient than the existing attribute reduction algorithms.

关 键 词：序决策信息系统 前向属性约简方法 优势粗糙集方法 属性约简 知识粒度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]