摘要
针对传统分类方法的缺陷,提出了一种基于AdaBoost-SVM的流量方法。该方法利用K-L变换从大量冗余流量特征中遴选出少量本征特征,有效降低了算法的处理复杂度;应用AdaBoost机制将一次分类过程等分成若干层基于支持向量机的弱分类器,使得分类方法简单、易于实现;通过分层组合和迭代权重的方法聚焦在困难分类的数据样本上,提高了分类器的准确性能。理论分析和实验结果表明:在降低计算复杂度的同时,Ada-Boost-SVM算法的准确性能够达到95%。
Aiming at the deficiencies of traditional classified methods, this paper presented a novel scheme called Intemet traf- fic classification based on AdaBoost-SVM. Herein, this method selected a few intrinsical flow'characteristics using K-L trans- form from a great deal redundant ones. In order to make the process easily implemented,AdaBoost equally partitioned the whole classification into several layers. It constructed one non-linear support vector machine in each layer. Through stratified combina- tions and iterativc weights, the algorithm focused on hard-classified data to improve the classifier' s performance. Theoretical a- nalysis and experimental results show that the algorithm based on AdaBoost-SVM can achieve the accuracy of 95% and better computational oerformance comoared with traditional K-means and NBC methods.
出处
《计算机应用研究》
CSCD
北大核心
2013年第5期1481-1485,共5页
Application Research of Computers
基金
国家"863"计划资助项目(2011AA01A103)
国家"973"计划资助项目(2012CB312901
2012CB312905)