并行PSVM算法及其在入侵检测中的应用(英文)

Parallel proximal support vector machine and its application in intrusion detection

基于并行PSVM(proximal support vector machine)分类法,利用ε-支持向量与原数据集等价的特点,将PSVM和cascade SVM模型高效结合,加速训练入侵数据集.提出一种新的PSVM增量学习方法,它能快捷更新分类器.通过大量基于著名的KDD CUP1999数据集实验,研究表明,该算法相对其他SVM方法,在保证较高检测率和较低误报率的同时,其训练时间降低80%,且能通过增量学习新数据集来有效更新分类器.

A novel training method based on parallel proximal support vector machine （PSVM） classification algorithm was proposed. The efficient PSVM and the cascade SVM architecture were used to reduce the time of training through the equivalence between the ε-support vectors and the original dataset. In addition, a new incremental learning method based on PSVM was used to make the update of the classifier easier. The experiments on the KDD CUP 1999 dataset demonstrate that the training time of our methods is 20% less than that of the other SVM methods under the condition of ensuring low false positive rate and high detection rate. it can update the classifier effectively by learning the characteristics of new dataset incrementally.