首页 | 官方网站   微博 | 高级检索  
     

连续属性空间上的规则学习算法
引用本文:权光日,刘文远,叶风,陈晓鹏.连续属性空间上的规则学习算法[J].软件学报,1999,10(11):1225-1232.
作者姓名:权光日  刘文远  叶风  陈晓鹏
作者单位:1. 哈尔滨工业大学威海分校,威海,264200
2. 哈尔滨工业大学计算机科学与工程系,哈尔滨,150001
基金项目:本文研究得到国家863高科技项目基金和煤炭科学基金资助.
摘    要:文章研究连续属性空间上的规则学习算法。首先简述了研究连续属性空间上的规则学习算法的目的和意义,并将规则学习理论中的一些基本概念推广到连续属性空间。在此基础上,研究了连续属性空间离散化问题,证明了属性空间最小离散化问题是NP困难问题,并将信息熵函数与无穷范数的概念应用到连续属性离散化问题,提出了基于信息熵的属性空间极小化算法。最后,提出了连续属性空间上的规则学习算法,并给出了数值实验结果。

关 键 词:规则学习算法  连续属性空间  信息熵  无穷范数  NP困难问题
收稿时间:5/4/1998 12:00:00 AM
修稿时间:1998/11/25 0:00:00

A Rule Learning Algorithm on Continuous Attributes Space
QUAN Guang-ri,LIU Wen-yuan,YE Feng and CHEN Xiao-peng.A Rule Learning Algorithm on Continuous Attributes Space[J].Journal of Software,1999,10(11):1225-1232.
Authors:QUAN Guang-ri  LIU Wen-yuan  YE Feng and CHEN Xiao-peng
Affiliation:QUAN Guang-ri; LIU Wen-yuan; YE Feng; CHEN Xiao-peng1(Weihai Campus Harbin Institute of Technology Weihai 264200)2(DePartment of ComPuter Science and Engineering Harbin Institute Of Technology Harbin 150001 )
Abstract:The rule learning algorithm on continuous attributes space is studied in this paper. First, thepurpose and the importance of studying rule learning algorithm on continuous attributes space are briefly introduced, and then some basic concepts in the theory of rule learning are extended to the continuous attributes space. On this basis, the authors study the problem to divide continuous attributes space, and prove that the problem of min dividing continuous attributes space is a NP hard problem. The concepts of information entropy and infinite normed apply to the problem of dividing continuous attribute space and a new algorithm of dividing continuous attribute space based on the function of information entropy are presented. At last, a rule learning algorithm on continuous attributes space is presented and the data results of the experiments are given.
Keywords:Rule learning algorithm  continuous attribute space  information entropy  infinite normed  NP hard problem  
本文献已被 CNKI 维普 万方数据 等数据库收录!
点击此处可从《软件学报》浏览原始摘要信息
点击此处可从《软件学报》下载全文
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-23

京公网安备 11010802026262号