基于TAN的网络流量分类方法

Internet traffic classification based on TAN

针对传统的基于传输层端口和基于特征码的流量分类技术准确率低、应用范围有限等缺点,提出了使用树扩展的贝叶斯分类器的方法,该方法利用网络流量的统计属性和基于统计理论的贝叶斯方法构建分类模型,并利用该模型对未知流量进行分类。实验分析了不同权值、不同规模的数据集对其性能的影响,并与NB、C4.5算法做了比较。实验结果表明,该方法具有较好的分类性能和较高的分类准确率。

Aimed at the problems of traditional port-based classification and payload-based classification,such as low accuracy and limited application region,tree augmented naive Bayesian classifier is proposed.The method builds a classification model using the network traffic statistical attributes and Bayesian method based on the statistical theory and classifies unknown traffic.The experiments compared with NB,C4.5,the effects of different sizes of dataset and variant weight on performance are analyzed.The results prove that the classifier has perfect classification capability and higher classification accuracy.