Short-Term Traffic Flow Forecasting of Urban Network Based on Dynamic STARIMA Model

This paper puts forward a hybrid spatio-temporal method of short-term traffic forecasting, i.e., Dynamic Space-Time Autoregressive Integrated Moving Average Model (Dynamic STARIMA). This method combines STARIMA model and Dynamic Turn Ratio Prediction model (DTRP) to enhance the forecasting performance and efficiency on urban intersections. To verify the Dynamic-STARIMA modeling method in real situation, an experimental model is constructed to produce forecasting traffic flow for part of urban network in Beijing, China based on actual data. The prediction accuracy of Dynamic STARIMA model is generally satisfying compared to other forecasting methods, which testifies the advantage and practicability of the proposed model.