Prediction of accidents severity at lighted intersections using sequential logit model

Traffic lights are considered as a way to improve traffic safety andperformance at intersections. Lots of crashes occur at intersections. There aremany accidents at intersections and these accidents cause financial losses,injuries, and even death. This research was carried out to predict the crashseverity of lighted intersections in Qazvin city, and examines the variables thataffect the incidence of severe accidents. Accident data were used base onQazvin police reports and data provided by Qazvin municipality traffictransportation department that includes 288 accidents at the 38 lightedintersections of the Qazvin city in the years 2014 to 2016. The accidents reportdescribes the circumstances in which an accident occurred. The accidents wereranked in three categories according to their severity including damage, injuryand fatal accidents that based on statistics (128 damages, 156 injury and 4fatalities). Sequential logit model was used in this research. The results showthat the presence of some variables such as accident time in the day, rate ofintersection volume and the existence of an island actually reduces thelikelihood of occurrence of a accident with a high severity. In contrast, thepresence of some variables such as accidents in summer, type of collision, howto collide, pedestrian traffic and the number of lanes, actually increases thelikelihood of accidents with a high severity.