Application of Artificial Neural Network Approaches for Predicting Accident Severity on Rural Roads (Case Study: Tehran-Qom and Tehran-Saveh Rural Roads)

Traffic accidents occur due to a combination of factors that lead casualties and injuries. By identifying the most effective factors, it is possible for safety authorities provide appropriate solutions decreasing accident severity implementing preventive measures. The aim this research was present models predict on two-lane two-way (TLTW) rural highways in Iran over one-year period from 2019 2020. Therefore, occurrence probability any type determined by artificial neural network (ANN)-based prediction using nine independent variables affecting severity. This study developed numerous ANN structures back-propagation model potential nonlinear relationship between accident-related factors. Results indicated among models, multilayer perceptron (MLPNN) with 6-2-2 partition had best performance power. standardized rescaling method covariates batch training. Also, 9, 5, 2 units were considered automatically input, hidden, output layers, respectively, hyperbolic tangent softmax used as an activation function hidden respectively. lowest cross-entropy error 39.6 highest correct percentage 82.5%, area under receiver operating characteristic (ROC) curve 0.852. Moreover, all variables, pavement condition index, roadside hazard, shoulder width, passing zone ratio greatest impact Finally, strategies proposed increase reduce along these roads.