Department of Civil Engineering,University of Tehran
Truck accidents are an issue of concern due to their severity. Logit modeling and Neural Network modeling are performed to investigate factors such as vehicle, roadway, environment and driver characteristics that can potentially contribute to the severity of truck accidents. The objective of this study is to present models that can predict the severity of truck accidents and to identify the important factors causing these accidents. Comparison between neural networks and logit modeling are made using vehicle crash data on two-lane rural highways in Iran. A variety of variables related to roadways, vehicles, environment and drivers, such as, driver fatigue, head-on collision and lack of vehicle control, are found to have a significant bearing on the severity of truck accidents. Also, investigating the marginal effects of variables showed the same variables to be significant. The results of the comparison between the logit and neural network model indicated that they both show similar patterns regarding the effects of different variables causing truck accidents, with the logit model providing better results.