Recursive Methods for Forecasting Short-term Traffic Flow Using Seasonal ARIMA Time Series Model.

Many Intelligent Transportation System (ITS) applications under the umbrella of Advanced Traffic Management Systems (ATMS) and Advanced Traveler Information Services (ATIS) call for the ability to anticipate future traffic conditions. Short-term traffic forecasting models play a central role in such applications. Previous research has shown that a three parameter SARIMA time series model is well suited for forecasting short-term freeway traffic flow. However, past application has been in a static form where the model has to be fitted separately for each location. This research implements the seasonal ARIMA model in a time-varying format imparting plug and play capability to the model. The properties of the SARIMA model for short-term traffic flow forecasting are discussed. Model sensitivity to the parameters is shown. Three different methods (Kalman filter, recursive least squares filter and least mean square filter) have been investigated for making the model adaptive. The stability and robustness of the SARIMA model has been demonstrated. Results show that all the three adaptive filters can be successfully used to make the model adaptive. The use of Kalman filter for practical implementation is recommended. Recommendations for further research in this regard are also presented.