Network Traffic Prediction Based On Support Vector Machines

Since the beginning of 1990s, Internet has achieved rapidly development, and Internet development is changing people's lives and traditional information industry framework. With the increase of network application and scale, the task of management becomes more and more heavily. The problem of network appears again and again. This is new challenge and assignment of network research. Traffic prediction has significant meanings for management, layout and design of large scale network. Traffic prediction with high quality is getting more and more important and exigent. Support vector machine (SVM) is a novel learning machine with many merits such as fast solving and strongly generalizing ability. The main research work is as follows:(1)The characteristic of network traffic which includes small time granularity and large time granularity is investigated. The sample selection and preprocessing are discussed. These conclusions offer reasons to the features selection of SVM and the modeling of traffic.(2)The model selection of SVW is studied, and the right kernel function is selected. The key to SVM constructions is to select appreciate kernel function. And RBF kernel function is selected by studying the primary properties of four kinds of common used kernels.(3)The parameter selection of SVM is studied, and some conclusions are gotten through experiments. LS-SVM has fast operating speed, so provides convenience of cross validation. Cross validation is used to select parameters, and experiment result shows satisfied effect.(4)The similarity and difference between SVM and neural network (NN) are investigated on theory foundation, method characteristic and model structure. The simulation results show that SVM has characteristics such as global optimum, strongly generalizing ability, short training time and simple operating compared with NN.