关系矩阵的知识粒度增量式属性约简  

Incremental Attribute Reduction based on Relational Matrix

作　　者：郑诚[1,2] 王波 洪彤彤 ZHENG Cheng;WANG Bo;HONG Tong-tong(Key Laboratory of Intelligent Computing ＆ Signal Processing （Anhui University） ,Ministry of Education,Hefei 230601 ,China） （ College of Computer Science and Technology,Anhui University, Hefei 230601, China)

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,合肥230601 [2]安徽大学计算机科学与技术学院,合肥230601

出　　处：《小型微型计算机系统》2018年第5期1000-1004,共5页Journal of Chinese Computer Systems

基　　金：安徽省高校自然科学基金重点项目(KJ2013A020)资助

摘　　要：由于现实中的数据集存在着大量的冗余属性,因此需要对它们进行属性约简.针对传统的属性约简算法不能很好的处理动态变化的数据集,本文运用关系矩阵的形式去表示信息系统的知识粒度,并研究了当数据集属性发生变化时,通过矩阵的视角来展示知识粒度的变化机制,根据这种机制可以快速地对知识粒度进行更新,从而提出了一种基于知识粒度的增量式属性约简算法,在UCI数据集的实验结果中,所提出的算法能够对动态变化的数据集选择出小而优的属性子集,并且有着较高的约简效率,从而验证了该算法的优越性.Because of data sets exist a lot of redundant attributes in reality,and they need to be reduced. For the algorithm of traditional attribute reduction can＇t deal with dynamic data sets,the form of the relationship matrix is used to represent the knowledge granularity of information system in this paper. Researching that when the data set attributes increase,the change mechanism of knowledge granularity is shown through the view of matrix. Then,the incremental attribute reduction algorithm based on the knowledge granularity is proposed according to the mechanism that can speed up to update knowledge granularity. The experimental results of UCI show that the proposed algorithm can select small and excellent attribute subset for dynamic change data sets,and has a higher accuracy of reduction.Therefore,the proposed algorithm＇s superiority is proved.

关 键 词：增量式更新 属性约简 关系矩阵 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]