决策树分类算法研究  

Research on Decision Tree Classification Algorithms

作　　者：张琳[1] 陈燕[1] 李桃迎[1] 牟向伟[1] 

机构地区：[1]大连海事大学交通运输管理学院,辽宁大连116026

出　　处：《计算机工程》2011年第13期66-67,70,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(70940008);高等学校博士学科点专项科研基金资助项目(200801510001)

摘　　要：ID3算法在选择分裂属性时偏向于选取属性取值较多的属性。针对该问题,引入属性重要性和属性取值数量2个参数对ID3算法的信息增益公式进行改进,从而提高取值数量少但较为关键的属性的重要性,使算法更好地反映实际决策情况,并根据凸函数的性质简化信息熵的计算,提高决策树的构造效率。通过实例介绍改进算法的具体应用方法,证明其性能相比原算法有所提高。ID3 algorithm tends to choose the attributes of more values as the splitting attributes.Aiming at the problem,this paper introduces two parameters including attribute importance and number of attribute values to improve the existed formula of information gain of ID3 algorithm.This contributes to enhancing the importance of the critical attributes with fewer values and making the algorithm better reflect the actual decision-making situation.According to the properties of the convex function,it simplifies the calculating formula of information entropy to improve the efficiency of constructing a decision tree.A concrete example is given to describe the specific application of improved algorithm,and the result shows that it is more efficient than the original algorithm

关 键 词：ID3算法 信息增益 属性重要性 属性取值数量 信息熵 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]