An IP Network Traffic Prediction Method based on ARIMA and N-BEATS

Abstract

We proposed and validated a novel method for predicting IP network traffic in this paper. The IP network traffic was forecasted by employing a combination of ARIMA (Auto-Regressive Integrated Moving Average) model and NBEATS model. It was mainly divided into the following steps: (1) predict the network traffic based on ARIMA model and calculate the prediction residual; (2) predict residual values of ARIMA model by N-BEATS model; (3) get the sum of ARIMA traffic prediction and the residual value predicted by NBEATS model as the final result. This method was validated on IP network traffic data from 31 provinces in China and the prediction performance was appraised by three metrics, Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), respectively. The experimental outcome showed that the method was superior to the traditional statistical methods (ARIMA) and classic machine learning methods (N-BEATS, LSTM and Prophet), and improved the effect of IP network traffic prediction.