Forecasting Network Traffic: A Comparison of Neural Networks and Linear Models

Abstracts of the 9th International Conference “Reliability and Statistics in Transportation and Communication” 2009
Irina Kļevecka

The main aim of the research was to produce the short-term forecasts of network traffic loads by means of non-linear neural networks and traditional linear methods such as ARIMA and exponential smoothing. In most cases the method of neural networks did not outperform linear models. Therefore, in contradiction to popular belief, the use of such complicated and time-consuming methods as neural networks is not always required.

Keywords
Telecommunications networks, traffic forecasting, neural networks, multilayer perceptron, ARIMA, exponential smoothing

Kļevecka, I. Forecasting Network Traffic: A Comparison of Neural Networks and Linear Models. In: Abstracts of the 9th International Conference “Reliability and Statistics in Transportation and Communication” , Latvia, Riga, 21-24 October, 2009. Riga: Transport and Telecommunication Institute, 2009, pp.36-36. ISBN 978-9984-818-22-1.

Publication language
English (en)