The Prediction Of Network Traffic Based On SVM And Resource Scheduling

With the continued rapid development of computers and the Internet, network is becoming more and more important in people’s life, people can no longer be satisfied by simple access to the Internet, the requirement of the people is higher and higher. A series of problems, such as network congestion, network failure and network security, always haunt us. So, the measurement, collection and forecasting of network has become one of the main challenges of network’s system operation.According to a large amount of data, network is a complex and multi-factor, the network traffic is clearly showing a high degree of self-similarity, time-varying and nonlinear characteristics, the traditional forecasting methods cannot achieve high accuracy, so, the traditional forecasting methods cannot achieve high accuracy. Support vector machine is a machine learning method, it is fast, and its generalization ability is strong, so in this dissertation, we use support vector machine (SVM) to predict.Support vector machine (SVM) is mainly based on the existing limited sample information, it balance the complexity of the model and the machine learning ability to get the best generalization ability. And it solves linear inseparable problem, through mapping it to a high-dimensional space by the kernel function.In this dissertation, after an accurate prediction of network traffic, CPU usage and memory usage are overall predicted. We design a fuzzy controller for urban petition docking platform, the fuzzy controller schedules resource based on forecast results, and we conduct experiments on the simulation platform, it has achieved good results.In this dissertation, the main research contents are as follows:1) The parameter selection of Support Vector Machine. The parameter has a huge impact on support vector machine’s modeling, the quality parameters directly affects the level of prediction accuracy. During graduate studies, I am concerned about a variety of new algorithms, and cuckoo search algorithm is applied to the parameter selection of support vector machines process Experiments compared with the existing algorithms, such as genetic algorithm (GA) and particle swarm optimization (PSO) algorithm, the cuckoo search algorithm obviously improves the efficiency and result accuracy of SVM;2) According to the records of network bandwidth, CPU usage, memory usage, in this dissertation, we use Support Vector Regression machine which based on the cuckoo search algorithm (CS-SVR) to predict, and schedule resources through the fuzzy controller designed by this article according to the results of the CS-SVR’s forecast, which making the utilization of the server’s resources to maximize, achieving load balance, and improving the quality of service.