Flight safety and reducing air accidents have always been the most important concerns of the aviation industry and related institutions. This study aimed to analyze descriptive and inferential statistics and predict the severity of air accidents using machine learning models. The research data included air accidents with variables such as flight phase, engine type, risk type, aircraft type, and number of engines. In the first stage, descriptive and inferential analyses showed that there is a significant relationship between the severity of the accident and technical and environmental factors. Then, logistic regression, decision tree, random forest, and robust boosted tree models were used to predict fatal accidents. The performance evaluation results showed that the XGBoost model provided the best results with an accuracy of 0.876 and a sensitivity of 0.676. This finding indicates the superiority of machine learning models over traditional statistical models in analyzing complex patterns of air accidents. The results of this research can be a basis for improving flight safety and reducing the risk of accidents in aviation organizations.