The pedestrian-ground injuries in pedestrian-vehicle collision have always been a research focus. This study explores the prediction model of pedestrian landing mechanism in pedestrian-vehicle collision. First, seven important parameters before, during and after the collision were extracted from 1300 cases. The obtained parameters are converted into input parameters through principal component analysis (PCA). Finally, the pedestrian landing mechanism prediction model under default parameters is constructed based on back-propagation neural network (BPNN), genetic algorithm (GA) optimised BPNN (GA-BPNN), support vector machine (SVM), decision tree (DT) and Random Forest (RF). The results showed that the GA-BPNN model is the optimal model under default parameters; the performance of GA-BPNN was improved after hyperparameter optimisation, and the prediction accuracy of the improved GA-BPNN (IGA-BPNN) model was 76.78%, 94.86% and 95.09%, respectively. Considering the pedestrian landing mechanism can significantly reduce the risk of vehicle braking control method and improve the protection efficiency.
Zou et al. (Wed,) studied this question.