Traffic accidents pose a significant global threat to public safety, with the World Health Organization (WHO) estimating that they claim the lives of approximately 1.25 million individuals each year. Without intervention, traffic accidents are projected to become the leading cause of death by 2030. Predicting accident severity and understanding their underlying causes represent crucial steps in developing effective strategies to prevent accidents and enhance overall traffic safety. This paper presented an in-depth analysis of accident severity prediction, considering a wide range of factors, including the vehicle, driver, environmental conditions, and more. The study aims to predict the extent of the severity of traffic accidents using a comprehensive dataset comprising over 4 million incidents that occurred across 49 states in the United States of America (USA) between February 2016 and December 2020. Various machine learning models, including logistic regression (LR), support vector machine (SVM), decision tree (DT), and random forest (RF), were implemented and rigorously evaluated against multiple performance metrics. The achieved results reveal that the RF model stands out with the highest accuracy of 91% in predicting accident severity. Additionally, this model demonstrates excellent performance across additional evaluation metrics, including a precision rate of 89%, a recall rate of 91%, a root mean square error (RMSE) of 18%, and an F1 score of 89%. These findings emphasize the exceptional predictive power and robustness of the RF model, making it a highly promising approach for real-world traffic accident scenarios. This research provides valuable insights into predicting accident severity, which is crucial for the development of effective accident prevention strategies and improvements in traffic safety.