A Traffic Interval Prediction Method Based on ARIMA

Abstract

The purpose of this paper is to construct an autoregressive integrated moving average model (ARIMA) for forecasting the traffic interval, which is helpful for the mobile industry to forecast the change of requirements for the peak of customer traffic, adjust the bandwidth dynamically, and improve the ability of active service. In this paper, the lateral time series analysis method is applied to analyze the data of the peak traffic of the uplink and downlink network from August to October of 2017 and 2018 to establish ARIMA prediction model for determining the parameters of it. MAPE method is used for model assessment and model diagnosis. Then, the optimal forecasting model is selected and the forecast error rate is calculated as the adjustment parameter of the forecast range. The model is used to forecast the customer flow range in the first three days of October 2019. Finally, the MIA method is proposed to compare with the LSTM algorithm and linear regression method for interval cumulative error comparison of traffic interval prediction. The result shows that the ARIMA (0,1,0) model has the lowest mean interval error rate of 15.5%, which proves the reliability of the model in predicting peak traffic interval.