首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于启发式信息熵的粗集数值属性离散化算法
引用本文:李春贵,王萌,原庆能.基于启发式信息熵的粗集数值属性离散化算法[J].广西科学院学报,2007,23(4):235-237.
作者姓名:李春贵  王萌  原庆能
作者单位:广西工学院计算机系,广西柳州,545006
基金项目:广西自然科学基金 , 广西教育厅科研项目
摘    要:在一致性假设前提下,以数据集的统计性质作为启发式知识,从候选离散点集中选择离散点,根据数据集的期望值和方差来确定搜索最优离散点的区域,提出一种新的基于信息熵粗集数值属性离散化算法,并采用UCI国际标准数据集来验证新算法.新算法与已报道的算法所得到的离散断点集完全一致,决策表的离散化结果也相同,但时间代价不同,新算法比其计算效率提高40%~50%.

关 键 词:信息熵  粗糙集  数值属性  离散化  统计性质
文章编号:1002-7378(2007)04-0235-03
收稿时间:2007-08-15
修稿时间:2007年8月15日

Discretization of Numerical Attributes in Rough Set Theory Based on Information Entropy with Heuristics Information
LI Chun-gui,WANG Meng and YUAN Qing-neng.Discretization of Numerical Attributes in Rough Set Theory Based on Information Entropy with Heuristics Information[J].Journal of Guangxi Academy of Sciences,2007,23(4):235-237.
Authors:LI Chun-gui  WANG Meng and YUAN Qing-neng
Institution:Department of Computer, Guangxi University of Technology, Liuzhou, Guangxi, 545006, China,Department of Computer, Guangxi University of Technology, Liuzhou, Guangxi, 545006, China and Department of Computer, Guangxi University of Technology, Liuzhou, Guangxi, 545006, China
Abstract:According to the consistency assumption in machine learning,the heuristics information of the data set statistic properties is used to select the discretization points from the candidate point set,in more detail,the mean and variance of data set are used to ascertain the region for searching optimal discretization points.A novel algorithm of numerical attributes discretization based on information entropy is proposed.The testing experiment with the UCI data sets has been performed.The results of experiment show that the discretization point set selected by using the new algorithm is the same as those by using the existing algorithm,and so does the results of decision tables discretization,but the time cost is different,the computing time of the new algorithm has saved about 40%~50% compared to the existing algorithm.
Keywords:information entropy  rough set  numerical attribute  discretization  statistic property
本文献已被 维普 万方数据 等数据库收录!
点击此处可从《广西科学院学报》浏览原始摘要信息
点击此处可从《广西科学院学报》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号