摘要
经典ID3算法在构造决策树有偏向于取值多的属性的缺点，主要原因是对概念的分级不够细致到位，造成信息熵的计算不准确，从而导致构造决策树时偏向于取值多的属性．本文提出了有效而简洁的分层概念的信息熵方法，在决策树的构造过程中较好地克服了ID3算法存在的缺点，最后通过一个实例验证了分层概念的信息熵方法的有效性性．
The conventional tree classification algorithm ID3 has the limitation of leaning towards more-value properties in the process of constructing a decision tree. The main reason is that the classification of concepts is not exact in ID3, which might lead to the incorrect calculation on information entropy,and cause the deflection on more-value properties in constructing the decision tree. The effectual and concise method of information entropy based on hierarchical concept is proposed in this paper, which can overcome the shortcoming of algorithm ID3 in the course of constructing decision tree. Finally the method of information entropy based on hierarchical concept is validated by an example.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2007年第B12期136-139,共4页
Acta Electronica Sinica
关键词
信息熵
分类器
概念分层
information entropy
classifier
concept hierarchy
