文档介绍:
改进的基于 SVM 决策树的多分类算法
刘靖雯,王小捷**
(北京邮电大学智能科学与技术中心,北京,100876)
5
10
15
20
25
摘要:标准的 SVM 是针对两类的分类问题,如何将两类问题推广到多类问题上,是目前研
究的一个热点。本文提出了一种改进的基于 SVM 决策树的多分类算法。该方法通过分析已
知类别样本的先验分布知识,根据新的类间可分性,把可分性最好的类划分放在父节点分类
器执行。通过采用非平衡树和平衡树结构,设计了新的非平衡 SVM 决策树和平衡 SVM 决
策树多分类算法。实验结果表明,该算法在不降低识别率的情况下,能大大减少系统的测试
时间,是一种有效的多分类算法,并在文本分类中获得良好效果。
关键词:自然语言处理;文本分类;支持向量机;决策树;多类分类器
中图分类号:TP181
Improved Multi-class Classification Algorithm Based on
SVM Decision Tree
Liu Jingwen, Wang Xiaojie
(Center for Intelligence Science and Technology, Beijing University of Posts and
munications, Beijing, 100876)
Abstract: Standard SVM is aimed at the problem of two class classification, how will the two class
problems are extended to multi-class problems, is currently a hot this paper a improved
multi-class classification algorithm based on SVM decision tree is proposed. The decision tree is
constructed based on the prior distribution of samples, which can make the more separable at the upper
node of the decision tree according to the new inter-class separability. This paper design two new
multi-class classification algorithms based on SVM decision tree for both unbalance and balance tree
structure. The experimental result indicates that the algorithm can significantly reduce system testing
time at the condition of not reducing identification rate, and is an effective multi-class clas