A Model for Predicting Schoolchildren Accidents in the Vicinity of Rural Roads based on Geometric Design and Traffic Conditions

awareness they gain from their surroundings. Recent statistics indicate that about 40 percent of road accident fatalities are pedestrians, 30 percent of which are under 18 years old. Based on the fact that almost two million Iranian students study in the vicinity of rural roads, this paper aims to develop a model for predicting the risk of students’ accidents near the aforementioned schools. Therefore, by gathering data from schools located in rural areas, schools are divided in three categories which are: no risk (no accidents/year), medium risk (one accident/year), and high risk (two or more accidents/year). A multinomial logit (MNL) model has been chosen and the utility function was considered as a combination of variables such as road width, functional speed, presence of school guardian, number of students and the ratio of average daily traffic to the distance of schools from roads. Results indicate that the proposed model can predict the risk of accident occurrence with the accuracy of more than 70 percent. Meanwhile, school guardian is known as an important variable in the prediction model. Also, results show the important role of the road width and proposed ADT/Dis variables.