期刊名称：International Journal of Future Generation Communication and Networking
印刷版ISSN：2233-7857
出版年度：2013
卷号：6
期号：6
页码：25-36
出版社：SERSC
摘要：In this paper an adaptive wavelet kernel based on density SVM approach for P2P traffic classification is presented. The model combines the multi-scale learning ability of wavelet kernel and the advantages of support vector machine. Mexican hat wavelet function is used to build SVM kernel function. The wavelet kernel function is tuned adaptively according to the density of samples around support vectors for several times during the training process. The experimental results show that the presented model can improve classification accuracy while reducing the number of support vectors and has better performance for solving P2P traffic classification
