基于粗糙集决策树优化研究

Variable precision multivariate decision tree based on rough set theory

决策树分类方法是一种有效的数据挖掘分类方法。单变量决策树结构简单,但规模较大。多变量决策树是为了进一步缩减树的规模而提出的决策树结构,通过选取属性的合理组合作为分裂属性,可使树的规模相对较小。文章在对以往所提出的混合变量决策树算法RSH2的抗噪性差和属性被多次选取等问题进行改进的基础上,提出了基于粗糙集的多变量决策树算法VPMDT。通过与ID3、HACRs、RSH2和C4.5等算法进行的实验比较表明,VPMDT有较好的时空性能,并保持较高的分类预测正确率。

The decision tree is an effective model in classification. The structure of univariate decision trees is simple while the magnitude is very large. However, multivariate decision trees can reduce the sizes of trees and maintain high prediction accuracy using the reasonable combination of several attributes as the split attributes properly. In this paper, an advanced multivariate decision tree algorithm named VPMDT（variable precision multivariate decision tree） is proposed based on the rough set theory to deal with the weaknesses of noise handling and attributes＇ multi-selecting. Extensive studies demonstrate that in comparison with state-of-the-art algorithms of ID3, HACRs, RSH2 and CA. 5, the VPMDT algorithm has better performance in the overheads of runtime and space as well as the prediction accuracy.