属性增加时基于矩阵方法的增量属性约简算法  

Matrix-based incremental attribute reduction method

作　　者：闫鑫 闫俊辉[2] Yan Xin;Yan Junhui(ShanXi University of Finance Economics, Faculty of Information Management, Taiyuan, Shanxi 030006, China;Yuncheng University, Department of Public Computer Teaching)

机构地区：[1]山西财经大学信息管理学院,山西太原030006 [2]运城学院公共计算机教学部

出　　处：《计算机时代》2018年第3期53-57,共5页Computer Era

摘　　要：随着通信和计算机技术的迅速发展,各行各业都积累了大量数据且这些数据集每时每刻都在动态变化,如何快速计算动态数据属性约简的问题是人工智能领域中的一个重要研究课题。针对决策信息系统属性增加时,如何快速计算变化后决策信息系统约简更新的问题,首先分析了计算变化后决策信息系统相对知识粒度和等价关系矩阵的增量更新机制,然后提出了决策信息系统属性增加后的增量属性约简方法,最后从公共数据网站下载了4组UCI数据集并对所提出的增量属性约简算法进行了仿真实验。实验结果表明了所提出的增量属性约简算法是有效的。With the fast development of computer network technology and communication technology,many real data in databases may vary dynamically.How to acquiring knowledge effectively and timely from the dynamic data has become an important problem.In this paper,the incremental mechanisms to calculate equivalent relation matrix and relative knowledge granularity based on matrices are introduced when multiple attributes are added into the decision system;The corresponding matrix-based incremental attribute reduction method is developed.Experiments on four data sets downloaded from UCI show that the proposed matrix-based incremental attribute reduction method is effective and efficient.

关 键 词：粗糙集 增量机制 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]