How does the Performance of the Data-driven Traffic Flow Forecasting Models deteriorate with Increasing Forecasting Horizon? An Extensive Approach Considering Statistical, Machine Learning and Deep Learning Models

The study on traffic flow forecasting models highlights the increasing importance of Intelligent Transportation Systems (ITS) in managing urban congestion, particularly as cities continue to grow. By analyzing real-world data from California's Harbor Freeway, the research assessed the effectiveness of statistical, machine learning, and deep learning models over 20 forecasting windows. The findings revealed that the ANFIS-GP model performed best in short-term predictions, achieving an RMSE of 0.038, while the Bi-LSTM model was more robust for medium-term forecasts, with an RMSE of 0.1863. This distinction is critical as it allows transportation planners to choose appropriate models based on their specific forecasting needs, ultimately aiding in the implementation of proactive traffic management strategies.