论文部分内容阅读
SVM决策树是解决多分类问题的有效方法之一,由于分类器组合策略不同,构成的决策树构型以及分类精确度也各有差异。提出基于欧氏距离的SVM决策树构造方法,通过两种欧氏距离组合策略,生成不同构型的SVM决策树。实验结果表明,采用组合策略二的SVM决策树分类器相比组合策略一,具有更高的分类精度和更短的训练及测试时间。
SVM decision tree is one of the effective methods to solve the multi-classification problem. Due to the different combination strategies of the classifiers, the structure of the decision tree and the classification accuracy are also different. A construction method of SVM decision tree based on Euclidean distance is proposed. Two kinds of SVM decision tree with different configurations are generated by two Euclidean distance combination strategies. The experimental results show that the SVM decision tree classifier using combinatorial strategy II has higher classification accuracy and shorter training and testing time than the combinatorial strategy.