Feature Selection And Classification Of Internet Video Traffic Based On Particle Swarm Optimization

With advances in Internet and streaming media technology,the network video traffic in the network traffic is continuous increasing.With the emerging of a variety of new network applications,the network environment has become increasingly complex.How to effectively allocate limited network bandwidth resources and ensure quality of network services becomes an urgent problem that ISP(Internet Service Provider)needs to solve.The technology of Internet video traffic classification enables the ISP to better allocate the network resources according to the different levels of QoS(Quality of Service)requirement of Internet video application.This thesis focuses on seven kinds of Internet video applications: standard definition video,high definition video,super clear video,live video,instant message video,P2 P video and HTTP video download.The main research work is as follows:A feature selection method based on ReliefF and Particle Swarm Optimization(PSO)is proposed.Firstly,it removes some irrelevant features to achieve the purpose of fast dimension reduction by Relief F;then continues to search optimal feature subset by using some better subset of ReliefF as initial PSO population and evaluating feature subset by inconsistency rate.ReliefF not only reduces the feature space,but also provides the priori knowledge of the PSO algorithm,thus improving the search efficiency and classification accuracy of the algorithm.Experiments show that the method can achieve better performance in different data sets than existing methods.How to find the statistical features combination which can reflect the nature of the Internet video is the key to the identification and classification of the Internet video traffic.By using the feature selection algorithm proposed in this paper,we can get some more discriminant Qo S statistical feature combinations,and verify it by feature map.The performance of the classification algorithm is related to the number of classes to be classified,in addition to the algorithm itself.And the performance of the classification algorithm will deteriorate when the number increases.Therefore,this paper designs a multi-layer SVM(Support Vector Machine)classification model for network video.Each SVM classifier is only used to identify a particular type of network video,and each uses different statistical features.Compared with existing methods,the multi-layer SVM cascade classification recognition algorithm proposed in this thesis has achieved better classification performance.