基于流量特征的用户互联网访问类型识别  

Traffic Features Based Categories Identification on Users' Network Behavior

作　　者：陈康[1] 黄晓宇[2,3] 陶彩霞[1] 关迎晖[1] 李磊[3] 王爱宝[4] 

机构地区：[1]中国电信股份有限公司广东研究院,广州510630 [2]华南理工大学经济与贸易学院,广州510006 [3]中山大学软件研究所,广州510275 [4]中国电信集团公司,北京100032

出　　处：《计算机系统应用》2014年第9期177-181,共5页Computer Systems & Applications

基　　金：国家高技术研究发展计划(863)(2012AA12A203)

摘　　要：近年来,对互联网用户在网络上的行为分析研究吸引了广泛的兴趣,分析的结果对网络运营商和普通用户都有重要的意义.研究用户在网络上的访问行为的类型识别问题,分析了一个由22万个网络数据包组成的数据集,从中提取统计特征,设计用户网络访问的类型识别算法,实验结果显示本文算法具有相当高的识别准确率.In recent years, the research on analyzing the users’ network behavior has attracted much attention. In this paper, we study the problem of identifying users’ network behavior categories. The research is based on a dataset that consists of 220 thousand network packets, with which we extract the statistical features needed for the identifications. We propose the identifying algorithm, and we also apply the algorithm to make categories identification on the network dataset. The results show the presented work can achieve very high accuracy.

关 键 词：网络流量 类型识别 特征选择 决策树 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]