An Empirical Evaluation of Multi-Step Prediction Performance

Traffic prediction constitutes a hot research topic of network metrology. MultiStep ahead prediction allows to predict more values in the future. Then, the result can be used to act proactively in many prediction applications. In this work, the AutoRegressive Integrated Moving Average (ARIMA) model and the linear minimum mean square error (LMMSE) are used for multiStep predicting. Via experimentation on real network traffic, we study the effect of some parameters on the prediction performance in terms of error such as the number of last observations of the throughput (i.e. lag) needed as inputs for the model, the data granularity, variance and packet size distribution. We also compared two multi-step prediction techniques: the Iterating Multi-Step technique (IMS) and the Direct Multi-Step technique (DMS). Besides, we performed a set of predictions based on packets size. Unexpectedly, we find that using more than two lags as inputs for the prediction model increases the prediction error. Using the last observation as the predicted value provides the same 1-step prediction performance as ARIMA or LMMSE model. The ARIMA model provides an acceptable multi-step prediction performance. Experimental results show that there is a granularity value at which the multi-step prediction is more accurate. They also show that the IMS technique provide more accurate traffic prediction than the DMS technique. We also find that the prediction of classified packets based on their size is possible. Especially, throughput of 1,500-byte packets is the less predictable.