基于强化正域的属性约简方法  

Attribute reduction method based on enhanced positive region

作　　者：史博文[1,2] 李国和[1,2,3] 吴卫江[1,2,3] 洪云峰 周晓明 

机构地区：[1]中国石油大学(北京)地球物理与信息工程学院,北京102249 [2]中国石油大学(北京)油气数据挖掘北京市重点实验室,北京102249 [3]石大兆信数字身份管理与物联网技术研究院,北京100029

出　　处：《计算机应用研究》2017年第1期107-109,127,共4页Application Research of Computers

基　　金：国家高新技术研究发展计划资助项目(2009AA062802);国家自然科学基金资助项目(60473125);中国石油(CNPC)石油科技中青年创新基金资助项目(05E7013);国家重大专项子课题(G5800-08-ZS-WX)

摘　　要：通过对多种约简方法进行比较,为了得到更好的结果,在传统基于属性依赖度的约简方法基础上,定义更精确的强化正域概念。通过对边界域的精确划分,得出各条件属性对决策属性的强化依赖度,并用自顶向下的启发式搜索算法得到约简结果。采用UCI标准数据集对基于强化正域约简方法 REPR进行测试,约简数据后构建的决策树规模小,分类精度高。实验结果表明,相比于经典方法,REPR能更有效地对决策表进行属性约简。Through a variety of reduction methods were compared, this paper proposed a more precise definition of the positive region based on the traditional attribute dependence in order to get better results. By dividing the boundary region accurately, the dependence of conditional attributes on decision attributes were enhanced, and obtained the results by using the top-down heuristic search algorithm. It used UCI standard data sets to test reduction method based on enhanced positive region REPR. And REPR can reduce the size of the decision tree and improve the classification accuracy, which shows that it can be more efficient for attribute reduction on decision tables than the other classical methods.

关 键 词：粗糙集 属性约简 正区域 属性依赖度 

分 类 号：TP301.5[自动化与计算机技术—计算机系统结构]