GARCH — non-linear time series model for traffic modeling and prediction

Forecasting of network traffic plays a very important role in many domains such as congestion control, adaptive applications, network management and traffic engineering. A good traffic model should have the ability to capture prominent traffic characteristics, such as long-range dependence (LRD), self-similarity and heavy-tailed distributions. In this paper, we propose a non-linear time series model, generalized autoregressive conditional heteroskedasticity (GARCH), with innovation process generalized to the class of heavy-tailed distributions. Our model is fitted on real data and our results confirms the goodness of fit of our model. We then evaluate a forecasting scheme based on our model. Comparative study with other generic models shows that our model have a better prediction accuracy. In addition, the parameter estimation is less complex than the other models used so far in modeling Internet traffic data.