C4.5算法的两点改进  

Two Improvements to C4.5 Algorithm

作　　者：乔增伟[1] 孙卫祥[2] 

机构地区：[1]江苏工业学院信息科学与工程学院,江苏常州213164 [2]上海交通大学振动冲击噪声国家重点实验室,上海200240

出　　处：《江苏工业学院学报》2008年第4期56-59,共4页Journal of Jiangsu Polytechnic University

摘　　要：C4.5作为一种重要的决策树算法尚存一些不足之处。针对C4.5对于连续属性最优分割阈值选择比较耗时的缺点,基于Fayyad边界点判定定理,提出一种改进最优阈值选择方法。针对C4.5不具备增量式学习能力的缺点,在改造树结构体的基础上,提出C4.5增量学习的改进方法。As an important decision tree algorithm, C4.5 still has two disadvantages. One is that it is very time--consuming to find the optimal threshold of continuous attribute. The other is that C4. 5 has no ability of incremental learning. Based on the boundary point theorem given by Fayyad, an improvement method of selecting the optimal threshold is proposed to overcome the first disadvantage. Also, after the modification of tree structure of C4.5, an incremental C4.5 algorithm is put forward to solve the second problem.

关 键 词：决策树 C4.5 最优阈值 增量式学习 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]