Busy/Idle Duration Model for WLAN Traffic and its Prediction Performance Using Autoregressive Method

摘要

This paper will study the busy/idle duration model and its prediction performance of autoregressive (AR) based predictor using the real environment data collected during rush-hour weekday evening at a railway station. The analysis shows that both busy and idle duration distribution largely appear as a generalized Pareto (GP) distribution with a different scale value. In addition, the scale value highly decides the prediction performance of the low-complexity linear AR based predictor. We also propose a new AR based predictor by separating busy/idle duration data into different streams to differentiate the scale value of the streams. The prediction performance of the proposed predictor can be improved for the streams with small scale value.