摘要： ID3算法采用一种对属性进行逐层的搜索和比较的“贪婪算法思想”。基于ID3算法的层间不相关性，该文考虑了生成树中相邻层的耦合，提出了一种改进的ID3的决策树算法(E-ID3)，E-ID3算法使用一种基于“统计出局部最优”的方法，获得比较好的启发式函数算法，并分析了E-ID3“算两步，走一步”的思想。实验证明，该优化算法对于构建决策树具有很好的效率。

关键词: 决策树, ID3, E-ID3, 加权熵

Abstract: Traditionally, the algorithm of ID3 is a “greedy” algorithm which searches and compares the attribute of each level in the decision tree. Based on the unrelativity among different levels, this article brings out an improved algorithm called E-ID3. It use a method of “compute two steps and exectue one step”, which can give a good suggestion. Experimental results show it has better efficiency than ID3 when building up decision tree.

Key words: decision tree, ID3, E-ID3, weighted entropy

中图分类号: 

• TP18