Spatio-Temporal Analysis and Severity Analysis using Machine Learning Classifiers for Electric Vehicle Crashes Data of Metro Manila, Philippines

Abstract

This study examines the factors affecting electric vehicle (EV) crash severity in Metro Manila by utilizing spatio-temporal analysis tools and machine learning classifiers, such as Random Forest (RF), K-Nearest Neighbors (KNN), Naïve-Bayes (NB), Artificial Neural Network (ANN), and an experimental Random Forest Classifier with Genetic Algorithm tuning (RF-GA). The spatio-temporal analysis reveals that most EV crashes occur during off-peak hours and are concentrated in areas with high population density. Crash hotspots were identified in locations without dedicated infrastructure. The results indicate that the NB and RF-GA models outperform the other classifiers in predicting crash severity, with car involvement as the most important variable. These findings hold significant implications for developing targeted interventions, identifying high-risk areas, and implementing measures to reduce the severity of EV crashes in Metro Manila, Philippines.