Prioritization of infrastructure factors affecting on safety of two-lane roads using preventative and reactive methods (Case study: Ahar-Tabriz road)

The present study is firstly aimed to identify, and prioritize factors affecting accidents on rural road of Ahar-Tabriz using artificial neural network (ANN), TOPSIS and multinomial logistic regression (MNL) models. Secondly, this study compares the results of priorities in three models based on preventive and reactive methods. The results of ANN model indicated that this model predicts accident severity via 86% in which prioritized factors such as horizontal and vertical curves, percentage of heavy vehicles, pavement condition, road drainage condition, volume of passing cars and use of speed control cameras, respectively. However, TOPSIS model has shown that the priority of the infrastructure factors affecting road safety include drainage condition, pavement condition, use of speed control cameras, horizontal and vertical curves, road signs, percent of heavy cars, road lighting condition, vehicle volume, and traffic calming, respectively. In addition, MNL model indicated that this model is capable of prediction in accident severity via 74.82% in which pavement condition, use of speed control cameras, road lighting condition, vehicle volume, and road signs are ranked as the most effective factors involved with accidents on rural roads, respectively. Therefore, by making a performance comparison based on Spearman's rank correlation coefficient, T-test analysis, it is found that there is no difference ANN and MNL models, however, there is a significant difference between TOPSIS and other models.