Minimax Probability Machine Regression for wireless traffic short term forecasting

Abstract

Traffic can reflect the latent rules and characteristics of the wireless network. Through researching, we found that the more accurate traffic prediction, the higher efficiency, utilization rate of network bandwidth and QoS can be guaranteed. Therefore, how to construct predictive models of wireless network traffic exactly is a major research topic. In this paper, Minimax Probability Machine Regression (MPMR) is proposed for forecasting wireless network traffic in 802.11 networks. Experiment provides the performance of the forecasting model and gives some comparative analysis. It evidences that the model is feasible. And compared with SVM, MPMR can not only obtain an efficient and satisfactory prediction efficiency but also less errors than SVM.