基于GA的多层次概念归纳学习方法  

A Multi-Level Concept Induction Method Based on GA

作　　者：刘明吉[1] 王秀峰[1] 饶一梅[1] 李宝林[1] 

机构地区：[1]南开大学计算机与系统科学系,天津300071

出　　处：《系统工程与电子技术》2001年第4期76-79,共4页Systems Engineering and Electronics

基　　金：天津市科学技术委员会资助课题! (0 0 370 0 111)

摘　　要：面向属性的归纳学习 (亦称概念提升 )是一种广泛使用的知识发现方法。通过归纳学习 ,使得属性域取值的抽象程度提高 ,从而得到较精练的数据集合 ,大大提高了规则的学习效率。但是实际应用环境中的数据属性维数非常多 ,属性概念层次也非常复杂 ,基于集合论的传统学习方法的效率变得越来越低。基于遗传算法的高搜索性能 ,提出了一个概念空间的特征概念层次优化搜索方法 ,特别是处理高维、具有复杂概念层次的问题时收到了较好的效果。Attribute - oriented induction (concept generalization) is a kind of KDD method widely used. Through concept induction we can improve the abstract level of attribute, thus we can get more succinct rule. But with the number of attribute increasing and the more and more complicated concept levels, the traditional method based on the set theory becomes lower and lower efficient. Based on the high searching ability and efficiency of GA, we propose a new heuristic algorithm in this paper, which seems to work well while dealing with large scale and complex problem, especially those with many dimensions and complicated multi - level concepts.

关 键 词：数据处理 多层次概念 归纳学习 遗传算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]