Abstrakt

Network traffic forecasting of ideological & political websites based on EMD_CPSO modified LS-SVM

Xia Xin, Li Hao

In order to enhance the accuracy of network traffic forecasting of ideological and political websites, in pertinence to the nonlinear and non-stable property of network traffic data, this paper introduces the chaos theory into PSO for optimization selection of the nuclear parameters and penalty coefficient of LS-SVM, and proposes an EMD_CPSO optimized LS-SVM (ECLS-SVM) network traffic forecasting model. It begins with EMD decomposition of the time sequence of network traffic data to extract detail features and trend features of network traffic data. Next, with the extracted network traffic features as the input into the CPSO optimized LS-SVM model, a network traffic forecasting model is built on the base of EMD_CPSO optimized LS-SVM. Finally, the network traffic data are utilized for a simulation experiment. By comparing the ECLS-SVM algorithm based single-step, 3-step, 5-step and 7-step forecasting results and the forecast time and mean square errors among different models, it is seen that the ECLS-SVM algorithm can effectively boost both accuracy and efficiency of network traffic forecasting, which attaches vital theoretical significance and practical value to instructing rational distribution and planning of network resources.