Network Video Service Feature Selection Based On Particle Swarm Optimization And Gravitational Search Algorithm

With the rapid development and popularization of the Internet,the video traffic that dominates the network traffic,is becoming more and more important.The identification and classification of network video services can help ISP(Internet Service Provirder)to optimize network resource management and configure traffic engineering reasonably,so as to guarantee the QoS(Quality of Service)of video service.In this thesis,seven kinds of network video services are identified,including live video,instant communication,P2 P video and HTTP video downloading,and online video streaming(including standard definition,high definition and ultra definition videos).The main research work is summarized as follows:A feature selection method based on Particle Swarm Optimization Gravitational and Search Algorithm,(PSOGSA)with variance importance coefficient(CI)is proposed.This method firstly sorts the feature subset according to the CI,to guide the initialization of the particle swarm with the top ranking features,and then combines the parameters of the PSOGSA with the selection of the features,for selecting the best feature subset with the lower complexity of CI as the evaluation function.Experiments show that this method can not only improve the accuracy of video traffic classification,but also reduce the time complexity of the algorithm.In addition,the classification of network video traffic has many problems,such as high dimensionality and large quantity of computations.Rendundancy exists among features,which will have a great impact on the choice of features and the performance of classifiers.Therefore,the network video traffic needs pre-processing to reduce the feature dimension,to save classification time and improve classification result.So,before selecting the feature,the feature subset is first sorted according to CI,to filter out some features,reduce the search dimension of the algorithm,and further improve the efficiency of the algorithm.The ultimate goal of feature selection is to find a statistical feature combination that can effectively distinguish among different kinds of network video services.We use the feature distribution graph to study the features of seven kinds of video services,and verify,by experiments,the effectiveness of the feature combination selected by the proposed method.