一种基于CART的决策树改进算法  

An Improved Algorithm Based on CART Decision

作　　者：宋广玲[1] 郝忠孝[1] 

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2009年第2期17-20,共4页Journal of Harbin University of Science and Technology

摘　　要：针对分析CART算法的要求被选择的属性是连续且有序的,并且只能产生两个子节点的问题,依据CART算法理论,采用A-CART算法,弥补了CART方法的不足.是一种选用离散无序的属性作为决策树的分裂节点,分裂后可以产生多个子节点的方法,并为此方法设计了分裂函数,通过一个实例给出了具体的使用方法.In view of basis on analyzing insufficient of the classifacation and regression trees （CART） algorithm, which requests the choosen attribute is continual and having order, and onlycan produce two children nodes, based on CART algorithm theory, this paper uses A - CART the algorithm, makes up the CART method insufficiency, namely the method which selects the discrete and disorder attribute to accomplish split point of decision tree, after splitting it may have many split node. It has designed a split function for the algorithm, and has given the concrete application method through an example.

关 键 词：决策树 CART 分裂函数 

分 类 号：TP311.12[自动化与计算机技术—计算机软件与理论]