基于改进SVM的网络流量分类方法研究.pdfVIP

学兔兔 第32卷 第7期 仪 器 仪 表 学 报 Vo1．32 No．7 2011年7月 Chinese Jourua1 of Scientific Instrument Ju1．2011 基于改进SVM的网络流量分类方法研究木 顾成杰。张顺颐 (南京邮电大学信息网络技术研究所 南京 210003) 摘 要：针对网络环境中存在大量噪声和网络流量中存在过多的冗余特征属性，提出了一种具有特征有效度的模糊支持向量 机(Fw—FSVM)，并将Fw—FSVM应用于网络流量分类领域。该方法根据不同样本点对分类贡献的大小赋予相应的模糊因子，可 以有效地消除噪声对分类精度的影响；同时计算网络流量中各个特征的有效度，消除弱特征属性或冗余特征属性对网络流量分 类精度的影响。实验结果表明，Fw．FSVM相比于其他网络流量分类方法能有效地提高网络流量分类精度且分类稳定性较高。 关键词：支持向量机；特征有效度；流量分类；分类精度 中图分类号：TP393 文献标识码：A 国家标准学科分类代码：510．5030 Network tralIfiC classification based on improved support vector machine Gu Chengjie，Zhang Shunyi (Institute of Information Network Technology，Nanjing University of Posts and Telecommunications，Nanfing 210003，China) Abstract：Since there are much noise and redundant features in network tramC，a novel fuzzy support vector machine with feature weighted degree(FW-FSVM)is proposed in this paper and applied in network traffic classification． Corresponding fuzzy factors are given according to the contribution of sample points to classification accuracy，which eliminates the influence of noise on classification accuracy．Feature weighted degree of each feature is also calculated to eliminate the influence of weak and redundant features on classification accuracy．Experimental results show that compared with other approaches of network traffic classification．FW—FSVM has higher classification accuracy and stability． Key words：support vector machine；feature weighted degree；traffic classification；classification accuracy 根据已知的应用层载荷的特征字符串来进行匹配，但不 1 引 言