The demand for transportation continues to expand along with economic development, making traffic safety an important public policy issue. This study focuses on the analysis of traffic accident reports from Budapest to address the crucial issue of traffic safety. By employing machine learning models, including Random Forest, BP Neural Network, KNN Classifier, and RBF Neural Network, we aim to predict the severity of accidents. Performance evaluation metrics, such as accuracy, recall, F1 score, ROC AUC, and precision, are utilized to assess the effectiveness of each model. The findings highlight the diverse effectiveness levels of the models, offering valuable insights into the factors that contribute to accident severity. Moreover, the study emphasizes the potential of machine learning in enhancing road safety. Notably, our analysis demonstrates that the KNN algorithm outperforms other models in accurately classifying the severity of accidents, underscoring its significance in understanding and improving road safety.