Traffic crash prediction based on the neuronal network approach: application to the ring way of paris
HAJ-SALEM ; BOUHELAL ; ZEGLAOUI
Type de document
COMMUNICATION AVEC ACTES INTERNATIONAL (ACTI)
Langue
anglais
Auteur
HAJ-SALEM ; BOUHELAL ; ZEGLAOUI
Résumé / Abstract
The paper focuses on the development of a Risk index model for traffic crash prediction, based on the application of a mixed approach: artificial neural networks, statistical analysis approaches. The inputs of the developed risk index model are the traffic measurements (volume and occupancy rate) and the calculated temporal left gradient. A global database including accidents and traffic measurements are used to validate the risk index model approach. The obtained results are promising while in some traffic conditions, the estimated risk index model is able to detect crash occurrence about 6 to 7 minutes prior to the crash time. This Risk index could be used as off-line safety evaluation index (evaluation process, off-line simulation) or real-time safety index monitoring for user information.
