摘要
ID3算法是一种典型的决策树归纳算法,它以信息增益作为选择扩展属性根结点的标准,并递归地生成决策树。但ID3算法倾向于选取属性取值较多的属性作为根结点,而且它假设训练集中各类别样例的比例应与实际问题领域里各类别样例的比例相同。提出一种新的基于粗糙集技术的决策树归纳算法,它是一种完全数据驱动的归纳算法,可以克服ID3算法的上述不足。
The ID3 algorithm is a typical decision tree induction method.hfformation gain measure is utilized to select optimal attributes with minimum entropy.Decision tree is recursively generated.However,there is natural bias in the information gain measure that favors attributes with many values over those with few values.Moreover,it assumes that the distribution of all classes' instances in the training set is same with the real problems.This paper presents a novel decision tree induction method,which is purely driven by the data used,and can overcome the drawbacks mentioned above.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第18期45-47,共3页
Computer Engineering and Applications
基金
河北省自然科学基金No.F2008000635
河北省应用基础研究重点项目No.08963522D
河北省教育厅首批百名优秀人才支持计划资助项目~~
关键词
决策树
ID3算法
粗糙集
上近似
下近似
decision tree
ID3 algorithm
rough sets
upper approximations
lower approximations