时频分析在P2P网络流分类中的应用研究  

Application of time-frequency analysis in classification of P2P flow

作　　者：张昕[1] 蒋晓路 于富财[1] 胡光岷[1] 

机构地区：[1]电子科技大学通信与信息工程学院,成都611731

出　　处：《计算机应用研究》2015年第10期3078-3082,共5页Application Research of Computers

基　　金：中央高校基本业务费资助项目(2672012ZYGX2012J006);留学回国人员科研启动基金资助项目

摘　　要：为了从众多网络流中识别出P2P流,以方便网络管理者更好地管理网络流量,提出了时频分析的思路,对网络流的时域特征进行时频分析,提取并形成时频特征参数,然后采用机器学习的方法对网络流进行分类。选取了短时傅里叶变换、小波包分解和Choi-Williams分布三种具有代表性的时频变换对网络流进行分析,交叉验证后的分类准确率大部分都在70%以上。实验结果表明,基于时频特征的分类方法具有较好的稳定性,是传统基于时域特征的分类方法的有效补充。In order to identify P2 P flow in numerous network flows,the network administrator can manage their traffic more effectively,this paper outlined some of the existing methods about P2 P flow classification,and put forward the time-frequency analysis method. This method analysed the result of time-frequency transform of the time domain features of network flow,and extracted the time-frequency features,and then classified the network flow based on the machine learning. This paper chose three representative methods： short time Fourier transform,wavelet packet decomposition and Choi-Williams distribution. After cross validation,the precision can mostly be more than 70%. The experimental results show that this method has a great improvement in classification accuracy compared with the traditional classification method based on the time domain feature.

关 键 词：流分类 时频分析 时频特征提取 机器学习 

分 类 号：TP393.09[自动化与计算机技术—计算机应用技术]