Owjimehr, Omid (University of Calgary), Demissie, Merkebe Getachew (University of Calgary, Calgary, Canada), Behjat, Laleh (University of Calgary)

Predicting Road Accidents Using Machine Learning Models

Scheduled for presentation during the Poster Session "Incident and emergency management" (WeBT17), Wednesday, September 25, 2024, 14:30−16:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 7, 2024

Abstract

Road travel accounts for most traffic accidents that have caused injuries, death, and property damage worldwide. Improvements in road traffic safety education, recent advancements in vehicle technology, and other environmental factors have decreased the number of road traffic accidents in developed nations. Many provincial and local governments envision the possibility of zero fatalities from road traffic accidents in the near future. Developing a proper accident prediction model to support such a vision is crucial. This study explores determinants of road collisions, emphasizing harsh winter weather. It then compares classical and Machine Learning models for collision prediction. Furthermore, it introduces the most influential factors in crashes concerning severe winter weather. All study parts are performed on the collision data in Calgary, Alberta, Canada, between 2017 and 2020. It is shown that all the weather attributes are correlated to collisions. It shows the importance of considering weather attributes in accident analysis and prediction. Based on the nature of the collision dataset, which is tabular and heterogeneous, Neural Networks showed higher performances than the other investigated models, with 92% accuracy. The developed models would allow transportation planners to apply these models for evidence-based policy implementation, including new speed limit recommendations.