In this study, text mining methods and machine learning algorithms were employed to analyze the final reports of Iranian air accidents during the period 2007 to 2025. The dataset consisted of 139 reports from the Civil Aviation Organization database, which were analyzed after textual preprocessing, including the removal of redundant characters, word normalization, stop-word elimination, and stemming. In the first stage, word frequency analysis and then topic modeling using the STM algorithm were applied to identify the main topics of the texts. The results indicated that the dominant topics included engine performance, approach and landing processes, aviation training, and helicopter accidents. Subsequently, accident severity was classified into four main categories using K-Nearest Neighbors, Support Vector Machines, Naive Bayes, Decision Tree, and Random Forest algorithms. Model evaluation showed that the Random Forest (RF) algorithm achieved the highest performance with an accuracy of 69 percent. These findings suggest that the integration of text mining techniques with machine learning algorithms can serve as an effective tool for uncovering hidden patterns in aviation accident reports and supporting safety-related decision-making. Accordingly, it is recommended that future research focus on developing more comprehensive and standardized accident report databases in order to provide a stronger foundation for conducting more complex and precise analyses. Also, using more advanced methods such as deep learning algorithms can further improve classification accuracy.