基于粗集理论和神经网络的集成化数据挖掘方法研究  

An Integrated Approach to Data Mining Based on Rough Set and Neural Network

作　　者：陆光辉[1] 肖人彬[1] 

机构地区：[1]华中科技大学CAD中心,湖北武汉430074

出　　处：《小型微型计算机系统》2002年第5期552-557,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金资助项目 (70 15 0 0 0 1);华中科技大学研究生研究基金资助项目

摘　　要：为了从大型数据库中获取有用的知识 ,本文提出了一种基于粗集理论和神经网络的集成化数据挖掘方法 .论文以所提出的研究框架为基础 ,首先给出了一种改进的粗集属性约简的算法和消除冗余属性的方法 ,进而采用面向对象的概念泛化进一步对数据库进行属性约简 ,最后用相似权值法得到产生式规则 ,并将所得规则用决策树来表示 .通过一个完整的应用实例演示了本文方法 。In order to obtain useful knowledge from large databases, an approach to data mining which integrates rough set and neural network is made in this paper. Based on the research framework, the paper presents an improved algorithm of attribute reduction and a method to remove the redundance attribute firstly. Then concept generalization which is derived from the object oriented idea is applied to attribute reduction. Finally, the similar weight approach is proposed to efficiently extract rules by utilizing the BP neural network, and the result is represented by decision trees. A practical example is given to illustrate the proposed method and verifies its effectiveness.

关 键 词：粗集理论 神经网络 集成化数据挖掘 数据库 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]