Network traffic prediction based on seasonal ARIMA model

Traffic prediction plays an important role in network layout, traffic management and etc. Two weeks network traffic of CERNET northwest center was investigated by seasonal ARIMA model and a traffic prediction model was proposed. Model parameters were educed by improved linear modeling. Traffic prediction data under different steps were computed according to the model. The experiments results show that the prediction data match real data approximately when prediction step is less than 10. The least mean square error of prediction is independent of time and only depends on step. The mean square error becomes bigger and the prediction effect becomes worse when the step becomes more.