基于层次分解的决策树  

A Decision Tree Based on Hiberarchy Decomposition

作　　者：杨宏伟[1] 赵明华[1] 孙娟[1] 王熙照[1] 

机构地区：[1]河北大学数学与计算机学院机器学习研究中心,保定071002

出　　处：《计算机工程与应用》2003年第23期108-110,175,共4页Computer Engineering and Applications

基　　金：河北省自然科学基金资助项目"基于模糊信息的示例学习理论及算法"(编号:698139)

摘　　要：目前人们经常使用决策树推理技术进行知识挖掘。以Quinlan1986年提出的ID3为代表的传统的决策树能较好地解决分类问题,但当类的个数增多时,所产生的单一决策树就会变得复杂,同时概括能力降低。该文采用基于层次分解的方法通过产生多层决策树来处理多类问题。与传统的单一决策树比较,基于层次分解的决策树在处理多类问题时有许多的优势。At present ,people often use decision tree reasoning technique to mine knowledge.The traditional decision tree,which is represented by ID3proposed by Quinlan in1986,can solve the classification problem well.But when the class number increase s,the produced single decision tree become s complex and generalization capability decreases.This paper uses the hiberarchy decomposition method to deal with multi-class problem by producing multi-level decision tree.Compared with the single decision tree,the decision tree based on hiberarchy decomposition has more advantages when dealing with the multi-class problem.

关 键 词：归纳学习 决策树 ID3 层次分解 知识挖掘 分类问题 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]