摘要
研究了SVM决策树分类器在文本分类中的应用,提出了一种有效的SVM决策树分类器的优化构建方法。该方法利用类间距离衡量两类间的可分性,并进一步用来描述各结点分类器类集合间的可分性。基于综合考虑结点分类器的类集合可分性,该方法能够获得优化的结点分类器类划分算法,由此构建的SVM决策树分类器在整体性能上得到优化,在文本分类中获得良好效果。
This paper proposes a new effective approach to optimize the SVM decision tree classifier while presents the research on text categorization using SVM decision tree classifier . In this approach, the within-class distance is used to measure the separability of two classes. Then the separability of class set of node classifier is measured based on that. Considering the separability of relative node classifiers , optimal class separation solution of each node classifier can be achieved. With the optimized class separation solution, the performance of SVM decision tree classifier is improved which has been tested by our text categorization experiment.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2005年第4期412-416,共5页
Pattern Recognition and Artificial Intelligence
关键词
文本分类
支持向量机
决策树
多类分类器
Text Categorization, Support Vector Machine, Decision Tree, Multi-Class Classifier
