基于属性差异与相似关系的知识约简方法  

Method for knowledge reduction based on difference and similitude of properties

作　　者：李文翔[1] 夏德麟[2] 

机构地区：[1]武汉科技大学信息科学与工程学院,湖北武汉430081 [2]华中科技大学交通科学与工程学院,湖北武汉430081

出　　处：《计算机工程与设计》2005年第12期3389-3391,3412,共4页Computer Engineering and Design

摘　　要：在数据挖掘系统的研究设计中,知识的获取和表示是一个备受关注的问题。提出的知识约简方法——差异相似矩阵算法,根据信息系统中各个样本的属性取值的差异性和相似性,构建矩阵模型,求取各决策类的最佳约简属性集,得出用于指导分类的规则知识。基于该算法开发的知识约简系统,能够有效地应用于大规模数据集的分析处理中。The retrieval and expression of knowledge is a more concerned problem in data mining research. A new method for knowledge reduction difference similitude matrix algorithm was introduced. According to this algorithm, a matrix was built by the difference and similitude of objects＇ property values, then the best property sets for reduction were acquired by analyzing the matrix elements, thus the rules for classification were brought forward. A knowledge reduction system was developed based on this algorithm, its efficiency and sueriority were proved by disposing large data source.

关 键 词：差异相似矩阵算法 知识约简 数据挖掘 UCI数据库 粗集理论 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]