基于支持向量机的流量分类方法

Traffic classification based on support vector machine

针对现有流量分类方法存在的准确率低、应用范围受限、计算复杂度高等问题,提出使用支持向量机方法来解决流量分类问题。使用公开的人工标注数据集作为训练集和测试集,通过有监督学习构建支持向量机流量分类器。此外,通过实验进一步分析了训练集大小、核函数、惩罚因子等因素对支持向量机分类性能的影响。实验结果表明支持向量机分类器可以达到98%以上的流分类准确率。

In order to solve the problems in current work, such as low accuracy, limited application region or high computation complexity, support vector machine （SVM） was applied to categorize traffic by application. The work capitalized on public hand-classified network dataset and used it to train and tested the supervised SVM traffic classifier. The improved accuracy of refined variants of this classifier was further illustrated, and the variants included the size of training dataset, kernel functions and penalty factors. The results indicate that it can achieve over 98% accuracy on per-flow classification with the SVM classifier.