一种基于修正信息增益的ID3算法  

A New ID3 Algorithm Based on Revised Information Gain

作　　者：张春丽[1] 张磊[1] 

机构地区：[1]电子科技大学计算机科学与工程学院计算智能实验室,四川成都610054

出　　处：《计算机工程与科学》2008年第11期46-47,94,共3页Computer Engineering & Science

摘　　要：ID3算法是决策树中影响最大的算法之一,它以信息增益为标准选择决策树的测试属性。这种算法存在不足之处,在选择合适的测试属性时,倾向于选择取值较多的属性,而在实际应用中,取值较多的属性未必是重要的。针对此算法的不足,本文提出了一种对增益修正的ID3算法,为改善ID3的多值偏向问题提供了一种有效途径。通过理论分析和实验证明,这种算法能较好地解决多值倾向的问题。The ID3 algorithm is a decision tree algorithm,which is important in the field of machine learning. The concept of information gain is proposed by Quinlan in the ID3 algorithm. Information gain is the selection criteria of the best splitting attribute for inducing decision trees. This algorithm has some drawbacks, one of which is that it tends to choose multi-value attribute as the best splitting attribute. However, the multi-value attribute is not necessarily important for classification in the real world. This paper presents a revised information gain of the ID3 algorithm in an attempt to solve this problem. From the theoretical analysis and experimental results we can see that the new method has a good effect on multivalue orientation of the ID3 algorithm.

关 键 词：ID3 决策树 信息增益 多值偏向 修正增益 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]