摘要
现有的决策树ID3、C4.5算法是一种快速有效的经典分类算法,但其有一个不足就是无回溯的自顶向下分析,造成所得的结果往往更多的是局部最优解而不一定是全局最优解。利用挖掘类比较技术,自底向上地分析描述,完善C4.5的分类算法,并实现自顶向下和自底向上共同分析,逼近全局最优解,取得了较好的效果。
The existing decision tree ID3 and C4.5 algorithm are fast and efficient classical classification algorithms, but there is a shortage that the non-backtracking top-down analysis, and the results tend to cause more of a local optimal solution and is not necessarily the global optimal solution. Compares the use of class mining technology, bottom-up analysis and improves C4.5 classification algorithm and to realizes top-down and bottom-up co-analysis, close to the global optimal solution, and achieves good results.
出处
《现代计算机》
2010年第1期38-41,共4页
Modern Computer
关键词
决策树
挖掘类
分裂属性
选择度量
Decision Tree
Mining Class
Split Attribute
Select Measure