粗糙集中一种连续属性预处理方法  

An algorithm for preprocessing continuous attributes in rough sets

作　　者：鄂旭[1,2] 高学东[1] 谢霖铨[1] 贺海钧[1] 

机构地区：[1]北京科技大学管理学院 [2]辽宁工学院计算机系,辽宁锦州121001

出　　处：《辽宁工程技术大学学报（自然科学版）》2005年第3期400-403,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：内蒙古自治区高等学校科学研究基金资助项目(NJ.02112)

摘　　要：针对在数据挖掘中,连续属性常常需要预处理问题,应用粗糙集理论对连续属性的不完备问题、离散问题进行了研究,提出了一种连续属性预处理方法。基于条件属性与决策属性间的对应关系完成了不完备数据的填补。依据划分区间的概念、连续属性离散化含义及其本质特征,定义了划分区间的加法运算法则,以此对填补后的信息表进行了划分区间运算,并以分类质量作为离散过程迭代约束条件,实现了信息表中连续属性的离散化。通过C++编写的算法进行数值示例及测试数据库,实验结果表明此算法有效可行。In data mining, continuous attributes sometimes need to be preprocessed. Based on rough set, the incomplete problem and the discretization problem are studied. And meanwhile a new algorithm for preprocessing continuous attributes is proposed. The incomplete data were filled up depending on the correlation between condition and decision attributes. According to the concept of demarcation and its essential, the paper defines a plus rule for the interval values. After adding interval values to each attribute with iterative constraints of classification quality, the continuous attributes were discretized. The illustration and experiment were done by the C++ program and the results indicate that the method is effective for preprocessing continuous attributes.

关 键 词：数据挖掘 粗糙集 预处理 划分区间 离散化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]