优势关系粗糙集增量属性约简算法  

Incremental Attribute Reduction Algorithm in Dominance-based Rough Set

作　　者：桑彬彬 杨留中 陈红梅[1,2] 王生武 SANG Bin-bin;YANG Liu-zhong;CHEN Hong-mei;WANG Sheng-wu(Department of Information Science and Technology,Southwest Jiaotong University,Chengdu 611756,China;Key Laboratory of Cloud Computing and Intelligent Technology,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学信息科学与技术学院,成都611756 [2]西南交通大学云计算与智能技术高效重点实验室,成都611756

出　　处：《计算机科学》2020年第8期137-143,共7页Computer Science

基　　金：国家自然科学基金(61572406,61976182);四川省国际科技创新合作重点项目(2019YFH0097)。

摘　　要：在现实生活中,数据不断累积增加,原有准则和决策之间的相互关系也随之动态变化,如何高效地计算属性约简是动态决策亟需解决的问题。增量更新方法可以有效地完成动态学习任务,因为它可以在原有知识的基础上获取新的知识。文中利用优势粗糙集方法研究了在有优势关系的数据中添加单个对象时的增量属性约简方法。首先,定义了优势集矩阵作为更新的目标,用来计算新的优势条件熵;其次,通过分析增加对象的3种不同情况,提出了优势条件熵的增量学习机制;然后,基于优势集矩阵设计了增量属性约简算法;最后,对6种不同的UCI数据集进行实验,用于比较增量和非增量算法的有效性和高效性。实验结果显示,提出的增量属性约简算法不仅在有效性上与非增量属性约简算法保持一致,而且在高效性上要远优于非增量属性约简算法。因此,所提算法能有效且高效地完成动态优势关系数据中属性约简的任务。In real life,as the data increase continuously,the relation between the original criteria and the decision making changes dynamically.How to effectively calculate the attribute reduction becomes an urgent problem to be solved in the dynamic decision making.Incremental updating method can effectively complete the dynamic learning task,because it can acquire new knowledge based on previous knowledge.This paper exploited the dominance-based rough set approach to study the incremental attribute reduction method when adding a single object to the data with dominant relation.Firstly,the dominant set matrix is defined as the target of the update to calculate the new dominant conditional entropy.Then,an incremental learning mechanism of the dominant conditional entropy is proposed by analyzing the three different situations of the adding object.After that,an incremental attribute reduction algorithm is designed based on the dominant set matrix.Finally,experiments on six different UCI data set are conducted to compare the effectiveness and efficiency of the incremental and non-incremental algorithms.The experimental results show that the incremental attribute reduction algorithm proposed in this paper is not only consistent with the non-incremental attribute reduction algorithm in term of effectiveness,but also far superior to the non-incremental attribute reduction algorithm in term of efficiency.Therefore,the proposed incremental algorithm can effectively and efficiently accomplish the task of attribute reduction in dynamical data with dominant relation.

关 键 词：优势粗糙集方法 属性约简 动态决策 增量学习 优势集矩阵 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]