二叉树多类SVM在网络入侵检测中的应用

Application of Binary Tree Multi-Class SVM to Network Intrusion Detection

本文在考察现有多类分类支持向量机(SVM)算法后,提出了一种基于二叉树结构的多分类器融合思想,融合过程充分考虑了类别之间的区分度,从而建立一颗相对优化的二叉树SVM的多类分类算法,并把改进后的多类SVM应用于入侵检测中以提高系统性能。在KDDCUP1999数据集上的实验结果表明了本方法的有效性。

Based on existing support vector machines（SVM） for multi-class classification, a novel multiple classifiers combination method based on binary tree structure is proposed.It considers discrimination degree between any two classes thus create a relatively optimized multi-class SVM algorithm with binary tree architecture.The improved method is applied to network intrusion detection.Experimental results carried out on KDD CUP 1999 data set indicate the effectiveness of the proposed method.