Traffic Prediction for Wireless Cellular System Based on Shrinkage Estimation

Abstract

In this paper, a traffic model is proposed based on shrinkage estimation with link load traffic data generated from the wireless cellular system. Compared with the traditional method, the spatiotemporal properties of different base stations (BSes) are considered, and a shrinkage estimation method Random Lasso is used to make variables selection, and to estimate the parameters of selected variables. The results show that the characteristics of traffic for the entire wireless cellular system can be captured effectively, and the prediction accuracy improves significantly. Besides, our research could be extended to other fields of spatiotemporal analysis with multivariate time series.