Research on Accident Prediction Models for Freeways in Mountainous and Rolling Areas

This study aims at predicting accidents for freeways in mountainous and rolling areas in China. The accident prediction model is established in line with the idea of IHSDM accident prediction model, and the geometric alignments are the main considerations during the modeling. Based on the horizontal and longitudinal alignments, the segments were firstly divided into subsections. For the subsections with unconstrained geometric alignments, the basic accident prediction model is established. The next step is to determine the geometric alignments that can significantly influence the occurrence of accidents. The BP neural network and sensitivity analysis are used, and the results show that the longitudinal slope, radius of horizontal curve, and the tangent length are the most significant ones. The modification factors of the three alignments are then determined. Combining the basic accident prediction model and the alignment modification factors, the final model is obtained. The application of the model shows that the relative error between model predictions and observed accidents is-5.85%~-7.87%.