结合粗糙集和禁忌搜索的网络流量特征选择  

Feature selection of network traffic using a rough set and tabu search

作　　者：顾成杰[1] 张顺颐[1] 杜安源 

机构地区：[1]南京邮电大学信息网络技术研究所,江苏南京210003 [2]中国移动通信集团设计院有限公司安徽分公司,安徽合肥230041

出　　处：《智能系统学报》2011年第3期254-260,共7页CAAI Transactions on Intelligent Systems

基　　金：国家"863"计划资助项目(2006AA01Z232;2009AA01Z212;2009AA01Z202);江苏省自然科学基金资助项目(BK2007603);江苏省高技术研究计划资助项目(BG2007045);江苏省重大科技支撑计划资助项目(BE2008134);江苏省科技成果转化专项基金资助项目(BA2007012)

摘　　要：针对网络流量特征属性的优化选择问题,提出了一种结合粗糙集和禁忌搜索的网络流量特征选择方法(RS-TS).该方法通过粗糙集算法对网络流量特征属性进行约简,将所得到的特征子集作为禁忌搜索的初始解,并利用禁忌搜索得到最优特征子集.实验验证RS-TS方法优于基于GA的特征选择方法和基于IG的特征选择方法,能够有效地去除网络流量的冗余特征属性,提高网络流量分类精度.A feature selection of network traffic using a rough set and tabu search （RS-TS） was proposed for the purpose of optimization in the feature selection of traffic classification. This approach reduced the traffic feature attribute with a rough set and established the feature subset as the initial value of a tabu search, as well as the optimal feature subset on the basis of a tabu search. The optimal feature subset with a tabu search can be selected on the basis of a feature subset. In contrast with the traditional feature selection methods based on GA and IG, RS-TS was vaXidated optimal by experimental results. It can diminish redundant feature attribution of network traffic effectively and greatly improve the classification accuracy.

关 键 词：粗糙集 禁忌搜索 特征选择 网络流量 

分 类 号：TP393.06[自动化与计算机技术—计算机应用技术]