基于粒化单调的不完备混合型数据增量式属性约简算法  

INCREMENTAL ATTRIBUTE REDUCTION ALGORITHM FOR INCOMPLETE MIXED DATA BASED ON GRANULATION MONOTONY

作　　者：张雨新[1] 孙达明 李飞[2] Zhang Yuxin;Sun Daming;Li Fei(Department of Information Engineering,Tangshan Polytechnic College,Tangshan 063299,Hebei,China;Computing Center,Northeastern University,Qinhuangdao 066004,Hebei,China)

机构地区：[1]唐山工业职业技术学院信息工程系,河北唐山063299 [2]东北大学计算中心,河北秦皇岛066004

出　　处：《计算机应用与软件》2021年第3期279-286,共8页Computer Applications and Software

基　　金：河北省教育厅创新创业课题项目(2017CXCY297)。

摘　　要：增量式属性约简是一种针对动态数据集的新型属性约简方法。然而目前的增量式属性约简很少有对不完备混合型的信息系统进行研究。针对这类问题提出一种属性增加时的增量式属性约简算法。在不完备混合型信息系统下引入邻域容差关系。基于邻域容差关系的粒化单调性,提出信息系统属性增加时邻域容差条件熵的增量式更新方法,并提出了不完备混合型信息系统下的邻域容差条件熵增量式属性约简算法。实验分析表明了该算法的有效性。Incremental attribute reduction is a new formal attribute reduction method for dynamic data sets.However,there are few studies on incomplete hybrid information systems for incremental attribute reduction.To solve these problems,an incremental attribute reduction algorithm with increasing attributes is proposed.In this paper,the neighborhood tolerance relation was introduced in incomplete hybrid information system.Based on the granulation monotony of neighborhood tolerance relation,an incremental updating method of neighborhood tolerance condition entropy was proposed when the attribute of information system was increased,and an incremental attribute reduction algorithm based on neighborhood tolerance condition entropy was proposed for incomplete hybrid information systems.The experimental results show the effectiveness of the proposed algorithm.

关 键 词：粗糙集 粒计算 属性约简 动态数据集 增量式学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]