Predicting traffic incident duration is a major challenge for many traffic centres around the world. Most research studies focus on predicting the incident duration on motorways rather than arterial roads, due to a high network complexity and lack of data. In this work we propose a bi-level framework for predicting the accident duration on arterial road networks in Sydney, based on operational requirements of incident clearance target which is less than 45 minutes.

Methodology

Using incident baseline information, we first deploy a classification method using various ensemble tree models in order to predict whether a new incident will be cleared in less than 45min or not. If the incident was classified as short-term, then various regression models are developed for predicting the actual incident duration in minutes by incorporating various traffic flow features and run intensive hyper-parameter tuning through randomized search and cross-validation.

Outcomes

• We identified GBDT and XGBoost as the best performing classifiers.
• We incorporated traffic flow data into the incident prediction by considering 5 different scenarios:
  • using all traffic flow in the subnetwork both in real time (closest to the time when the accident has been signalised) and 1 hour prior.
  • Using 5 closest sections affected by incidents and their reported traffic flows.
  • Aggregated and averaged traffic flow in the vicinity of the accident.
  • Fluctuating radius of 100m, 200 meters,..600 meters around the incident locations.
• XGBoost outperformed all repressors by almost 53%,
• The affected number of lanes and the hour fo the day are the most 2 important features which can affect the exact prediction of how long a fresh accident will last.

Accepted Paper

Mihaita, A.S., Liu, Z., Cai, C., Rizoiu, M.A “Arterial incident duration prediction using a bi-level framework of extreme gradient-tree boosting”, ITS World Congress 2019, Singapore, 21-25 Oct 2019, Preprint link