基于修正系数的决策树分类算法  

Decision tree classification algorithm based on modified degree

作　　者：狄文辉[1] 李卿[2] 楼新远[2] 

机构地区：[1]河南机电高等专科学校计算机科学与技术系,河南新乡453002 [2]西南交通大学信息科学与技术学院,四川成都610031

出　　处：《计算机工程与设计》2008年第24期6344-6346,共3页Computer Engineering and Design

基　　金：河南省基础与前沿技术研究计划基金项目(082300410390);河南省高等学校青年骨干教师计划基金项目(2005-174);河南省教育厅自然科学研究基金项目(2007520008)

摘　　要：ID3算法是决策树算法中的经典算法,但存在多值偏向问题。一些改进的ID3算法虽避免了多值偏向问题,但多存在主观性强、没有考虑属性信息熵等问题。为了解决该问题,提出了一种基于修正系数的决策树分类算法MC。该算法利用修正系数降低取值个数多的属性的信息增益,并通过实验与ID3算法进行了比较,结果表明,当样本集中各属性取值个数不同时,算法MC在生成决策树的结点总数和分类准确率上明显优于ID3算法。ID3 is the key algorithm of decision tree algorithm, but there is existing variety bias problem. The shortcoming of current algorithms for improved ID3 which are subjective to measure the attributes which are important or not and without regard to attributes information entropy is analyzed. For solving these problems, a decision tree algorithm based on modified degree is proposed. The main idea of the approach algorithm is using modified coefficient to reduce information gain of attributes which has many values. Experiments compared to ID3 algorithm show that the improved algorithm in term of the accuracy rate between decision trees and the classification is superior to ID3 when the samples of the value of the number of different attributes.

关 键 词：决策树 多值偏向 修正系数 ID3算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]