作　　者： (鞠恒荣）; (李华雄）; (周献中）; (黄兵）; (杨习贝）;

机构地区： 南京大学工程管理学院,南京210093

出　　处： 《计算机科学》 2017年第9期34-39,57,共7页

摘　　要： 序贯三支决策是三支决策理论近年发展起来的一种新型决策方法。传统的序贯三支决策方法鲜有针对序贯信息粒的构建和其在分类学习中的应用的研究。针对这两个问题,研究了Local约简与Global约简之间的内在序贯性,并以此构建了具有约简特性的序贯信息粒。在此基础上设计了一种序贯三支分类器。实验结果表明,该序贯三支分类器不仅能很好地在合适信息粒上进行分类,而且较传统的分类算法提高了数据集的分类精度。 Sequential three-way decision is a novel decision approach of three-way decision theory in recent years. Howe ver,in classical sequential three-way decision research,little attention was paid to two important issues, one is the con- struction of sequential information granule,and the other is the application in classification learning. To address such is sues,the intrinsic sequential properties of local and global reductions were studied firstly in this paper. Based on such properties, the sequential information granule was constructed with reduct＇s property. Furthermore, a sequential three- way classifier was proposed and designed. The experimental results show that, the proposed classifier is not only good at making classification at an appropriate information granule, but it can also improves the classification accuracy when compared with several classical classifiers.

关 键 词： 分类器 约简 序贯 粗糙集 三支决策