Prediction of Long-Range Dependent Processes for Teletraffic Applications

Recent studies have shown that trafic in telecommunication networks exhibits long-range dependence (LRD). Accurate modelling and analysis of teletrafic incorporating LRD is therefore required in network engineering. Prediction of trafic levels can play an important role in teletrafic analysis for dynamic resource allocation and traffic management. This paper presents the formulation of a model based recursive , linear minimum mean-square error predic- tor for LRD processes. A Kalman predictor is pro- posed for LAD processes modelled by fractional autoregressive integrated moving average CfARIMA) models. The family of fARIMA models can account for long range, as well as short range and qwi-peri- odic dependencies typical of teletraffic data.