Comparison of SARIMA, NARX and BPNN models in forecasting time series data of network traffic

The investigation and forecasting network traffic usage is an essential concern in the academic activities of university. This paper reports how to apply and compare SARIMA, NARX, and BPNN by using short-term time series datasets. The network traffic datasets are obtained from the ICT Universitas Mulawarman. As a result, the determination of several prediction models will continue to be an alternative for researchers to obtain more accurate prediction results. The first analysis used the SARIMA ((2,1,1)(2,1,2)12) with MSE of 0.064 indicated that it was a good model. The second analysis used the NARX models by using architecture 189∶31∶94 with performance value of MSE was 0.006717 respectively. The third one used the BPNN with two-hidden-layers (5-10-10-1) architecture with MSE value of 0.00942479. Finally, we compared the performance of methods using MSE. Based on the experiment, the artificial neural networks (ANN) i.e., NARX and BPNN models have been successfully to support the time series datasets in order to predict the future.