Abstract
Ensuring public safety on our roads is a top priority, and the prevalence of road accidents is a major concern. Fortunately, advances in machine learning allow us to use data to predict and prevent such incidents. Our study delves into the development and implementation of machine learning techniques for predicting road accidents, using rich datasets from Catalonia and Toronto Fatal Collision. Our comprehensive research reveals that ensemble learning methods outperform other models in most prediction tasks, while Decision Tree and K-NN exhibit poor performance. Additionally, our findings highlight the complexity involved in predicting various aspects of crashes, as the Stacking Regressor shows variability in its performance across different target variables. Overall, our study provides valuable insights that can significantly contribute to ongoing efforts to reduce accidents and their consequences by enabling more accurate predictions.