基于决策树的网络流量分类方法  

Traffic Classification Based on Decision Tree

作　　者：于孝美 陈贞翔[1,2] 彭立志[1,2] 

机构地区：[1]山东省网络环境智能计算技术重点实验室,山东济南250022 [2]济南大学信息科学与工程学院,山东济南250022

出　　处：《济南大学学报（自然科学版）》2012年第3期291-295,共5页Journal of University of Jinan(Science and Technology)

基　　金：国家自然科学基金(60903176);山东省中青年科学家奖励基金(BS2009DX037)

摘　　要：针对传统流量分类方法(基于端口和有效载荷)分类不可靠的问题,提出基于C4.5决策树算法,根据训练集中属性的信息增益比率构建分类模型,按属性对测试数据集进行预测,通过查找分类模型实现对网络流量的分类。在公开数据集和自己采集的数据集上进行实验,结果表明,采用C4.5决策树算法对网络流量分类,平均分类精度为93%,单类别分类精度均在90%以上,能有效地实现对网络流量应用类型的识别。Aiming at the problem of instability in traditional traffic classification methods, a traffic classification method based on CA. 5 decision tree is proposed, which establishes models on the information gain ratio from the training set. Classifier is tested by attrib- utes on test dataset, as well as network traffic is classified by searching classification models. Experiments show that the overall accuracy of our method achieves more than 93% ,and the accuracy of single class is more than 90% on open dataset. So the method is effective for classifying various kinds of traffic.

关 键 词：流量分类 决策树 网络流 统计属性 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]