一种基于凸壳算法的SVM集成方法

为提高支持向量机(SVM)集成的训练速度，提出一种基于凸壳算法的SVM集成方法，得到训练集各类数据的壳向量，将其作为基分类器的训练集，并采用Bagging策略集成各个SVM。在训练过程中，通过抛弃性能较差的基分类器，进一步提高集成分类精度。将该方法用于3组数据，实验结果表明，SVM集成的训练和分类速度平均分别提高了266％和25％。

SVM Ensemble Approach Based on Convex-hull Algorithm

To improve the training speed of Support Vector Machine(SVM) ensemble, this paper proposes a new approach of SVM ensemble using convex-hull algorithm. The approach applies convex-hull algorithm to get from each class the hull vectors and takes these hull vectors as the training dataset for every base-classifier, Bagging method is used to aggregate the base-classifiers. Threshold is set to discard the base-classifiers with weak performance in training the ensemble to further improve the classification accuracy. Experimental results obtained from applying the proposed approach to 3 different datasets indicate that on average it accelerates training by 266% and speeds up classifying by 25%.