A new SVM decision tree
Binghan Liu · Fuzhou daxue xuebao. Ziran kexue ban · 2007
SVM has good generalization performance when the number of training samples is very small and the dimension of feature space is very high.But it is not suitable for multi-class classification.This paper analyzes the basic SVM and the SVM classifier multi-class classification,especially about the SVM decision tree,then proposes a method for partition of the set of classes on each node classifier to build up SVM decision tree.The results of experiment demonstrate that the SVM decision tree built up by this method has a good classification performance.