多变量决策树的研究与探讨
首发时间:2006-12-20
摘要:目前大多数决策树构造方法在每个节点上只检验单一属性,这种单变量决策树忽视了信息系统中广泛存在的属性间的关联作用,而且修剪时往往代价很大。并且现有的多变量决策树构造方法,在数据间存在明显的非线性成分时,其构造的树的精度不够高。本文提出了一种基于核函数的非线性主成分分析(KPCA)多变量决策树构造方法,提取信息系统中的若干主成分来构造多变量决策树。理论表明,这是一种操作简单,效率很高的决策树生成方法.
关键词: 数据挖掘;单变量决策树;多变量决策树; 基于核函数的非线性主成分分析
For information in English, please click here
Research on Multivariate Decision Tree
Abstract: Now, most decision tree construction algorithms check up only one attribute on each node. This class of decision tree, called unvaried decision tree, ignores the connection effect among the attributes inside the certain information system, which is actually widely occur. Further more, the cost of pruning is usually large. Currently, approach for multivariate decision tree construction is not efficient, when nonlinear correlations are contained between data. So kernel principal component analysis-based approach for multivariate decision tree construction is proposed in this paper. And several principal components should be extracted from the information system to constructing decision tree. The theoretic demonstrate it is a simple decision tree construction algorithm with high efficiency.
Keywords: Data Mining Unvaried Decision Tree Multivariate Decision Tree Kernel Principal Component Analysis
基金:
论文图表:
引用
No.1040189543116657****
同行评议
共计0人参与
勘误表
多变量决策树的研究与探讨
评论
全部评论0/1000