A new ensemble learning with support vector machines

Cascade of classifiers can, in general, improve the performance of any given classifier. In this paper, we present a new cascade classifier constructed with the support vector machine (SVM) classifiers where a set of SVMs is learned repeatedly with the bounded support vectors of the previous SVM. A binary decision tree is formed using the learned classifiers to take the decision of a new example. Experimental results show that the proposed method can improve the generalization performance over a single SVM.