Abstract: In classical ID3 algorithm, “information plus” is regarded as illuminative information in selecting examing attribute. But it has the insufficiencies. In order to solve it, this paper brings forward cognitive entropy concept and information compensation concept, makes thinks of compensative information of the attribute as illuminative information, and designs a kind of CID3 algorithm based on compensative information. Aanalysts a great number of instances testify that CID3 algorithm has much more priority than ID3 algorithm.

Key words: decision tree, compensative information, cognition entropy, ID3 algorithm, CID3 algorithm

摘要： 在经典的ID3算法中，用“信息增益”作为选取检测属性启发信息，为了解决该算法存在的不足，该文提出了认识熵和信息补偿的概念，将属性的信息补偿量作为启发信息，设计了一个基于信息补偿量的CID3算法。实例分析表明，CID3算法比ID3算法更具有合理性。

关键词: 决策树, 信息补偿量, 认识熵, ID3算法, CID3算法

CLC Number: 

• TP18