摘要
为了将传统的决策树无法管理的、由各种分类算法所发现的大量的有意义的规则进行有效的存储、剪裁和使用 ,提出了广义决策树结构。它将传统决策树的结构进行扩展 ,能够以较少的存储代价管理所发现的所有分类规则 ,且易于表达规则之间的关系。提出了有效的优化策略。以此树为基础 ,将决策树分类算法与基于关联规则的分类算法进行了概括统一 ,并提出了相应的算法。实验结果证明 ,广义决策树克服了传统决策树的缺点 ,并且适宜于维护。
A generalized decision tree was developed to effectively store, prune and use large amounts of meaningful rules used by various classification algorithms which can not be managed by traditional decision trees. The system extends the structure of traditional decision trees so it can store all the classification rules found with less storage cost and can more easily express the relationships between rules. An effective optimization strategy was developed to speed up the rule search process. The structure can generalize and unify decision tree classifications and classifications based on association rules. Test results show that the generalized decision tree overcomes the weaknesses of traditional decision trees and can be easily maintained, pruned and searched.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2004年第6期762-765,777,共5页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目 ( 70 2 73 0 5 5 )
国家自然科学基金创新研究群体科学基金项目( 70 3 2 10 0 1)
关键词
信息处理
数据挖掘
分类
决策树
关联规则
information processing
data mining
classification
decision tree
association rule
