Multi-period forecasting of accident duration based on multiple sources of real-time traffic data

The aim of this study is to build the adaptive ANN-based models for accident duration prediction. Two types of accident duration models are conducted. Model A is used to forecast the duration at the accident notification and Model B provides multi-period update of incident duration after accident notification. The APE and MAPE of forecasted accident duration at each time point of forecast are mostly under 20%. With these two models, the estimated duration can be provided by plugging in relevant traffic data as soon as the incident is notified. The traveler and traffic management unit can generally realize the impact by the forecasted accident duration. From the assessment of model effects, this study shows very promising practical applicability of the proposed models in the Intelligent Transportation Systems (ITS) context.