Predictive classification of pedestrian-vehicle crossing behaviors using a hybrid mountain gazelle optimizer-enhanced Long Short-Term Memory model

This paper addresses the safety issues of pedestrian-vehicle interactions at unsignalized pedestrian crossings by proposing a Hybrid Mountain Gazelle Optimizer-Long Short-Term Memory (HMGO-LSTM) model. The proposed model combines the Hybrid Mountain Gazelle Optimizer (HMGO) algorithm with a Long Short-Term Memory (LSTM) network, using HMGO as an LSTM hyperparameter optimizer. Real-world datasets of pedestrian and vehicle crossings from Ma'anshan were used to train and evaluate the model. The HMGO-LSTM model was compared with traditional classifiers such as K-Nearest Neighbors (KNN), Random Forest (RF), and Genetic Algorithm-Backpropagation (GA-BP). The results show that the HMGO-LSTM model outperforms these classifiers in predicting pedestrian-vehicle interaction behaviors, achieving higher classification accuracy and F1 score. The model also optimizes safety intervals for crossings, leading to new speed limit recommendations. Overall, the HMGO-LSTM model provides a robust theoretical foundation for managing and designing safer pedestrian and vehicle crossings.