This research presents a comprehensive analysis of traffic accident severity using machine learning (ML) models to predict outcomes and contribute to traffic safety enhancement. Various ML techniques, including Logistic Regression, Decision Trees, Random Forests, and Neural Networks, are implemented and rigorously tested on a substantial dataset predominantly featuring U.S. data. The study effectively handles the dataset, addressing missing values, outliers, and data transformation, showcasing technical proficiency and innovation in dealing with large-scale data. Despite achieving technical robustness and practical utility, the research encounters limitations due to the imbalanced nature of the dataset and the predominance of less severe accidents. The constraint of computational resources restricts the exploration of basic ML models and preprocessing techniques.