This study investigates the application of Support Vector Machine (SVM) models using various kernel functions to predict traffic accident severity, a critical task in improving road safety measures. The analysis focuses on evaluating the efficacy of these models in terms of classification accuracy and generalization capacity, essential for reliable prediction in real-world scenarios. Results indicate that the Radial Basis Function (RBF) kernel significantly outperforms the Linear, Polynomial, and Sigmoid kernels, demonstrating superior performance across both training and testing phases. These findings underscore the importance of selecting the appropriate kernel function to optimize SVM model performance, with the RBF kernel emerging as the preferred choice due to its ability to capture complex data patterns and facilitate accurate classification. The study’s insights have practical implications for enhancing predictive models in traffic safety, and future research should explore novel kernel functions and ensemble techniques to address challenges related to dataset heterogeneity and dimensionality. These contributions advance machine learning methodologies, particularly for classification tasks in diverse domains where accurate predictions are crucial.