SUGGESTIONS FOR THE SELECTION OF METHODS OF NETWORK TRAFFIC FORECASTING

The efficiency of methods of forecasting of intensity of the network traffic in various points of the telecommunication network is considered in the article. As basic data the real values received for 5-second and 5-minute monitoring intervals are used. The analysis of the obtained data indicates similar autocorrelated characteristics, thus for the traffic with 5-seconds monitoring interval sharp changes of intensity are more characteristic. It is shown that in some cases the considered models possess almost identical efficiency. It is established that for traffic forecasting in the core network the ARIMA model is more suitable, and for border points the artificial neural networks are more relevant.