基于Bagging支持向量机集成的入侵检测研究

Intrusion Detection Based on Support Vector Machine Ensemble with Bagging

对大数据集来说,支持向量机的时空耗费非常大,本文采用bagging技术对支持向量机进行集成。首先用bootstrap技术对训练样本集进行可重复采样,使所得到的新子样本集有较大差异,然后用多个支持向量机对各子样本集进行学习,并将学习后的结果用多数投票法集成最终的结论。实验表明,支持向量机集成对入侵检测数据有比单个支持向量机更好的分类性能。

For large data sets, the performance of Support Vector Machine (SVM) is not satisfied, because of its high complexity of time and space. In order to reduce the complexities, we propose a new method that uses the SVM ensembles with bagging (bootstrap aggregating) in this paper. We train each individual SVM independently using the randomly chosen training samples via a bootstrap technique. After that, they are collected to make a decision according to the majority voting. The experiment results for the intrusion detection data classification show that our proposed SVM ensemble with bagging outperforms any single SVM in terms of classification accuracy.