Due to the aviation accident is rarely predictable and often irreversible, how to ensure aviation safety is of uttermost importance. Textual aviation accident reports contain the cause and process of the accident which could help people understand incidents. However, the cause of the accident always is summarized by the expert and the accident report would be incomplete, the identification of aviation safety accident risk is not timely and accurate. In this paper, a safety risk identification model is proposed, aiming to identify the correlation between aviation safety accident risk factors by machine learning from textual aviation accident reports. In detail, the feature of aviation accidents is extracted and classified by text mining technology, on this basis, the correlation coefficient matrix between different features is established. Finally, the correlation network of aviation safety risk is proposed, and the risk propagation process of accidents is developed based on the network to identify aviation safety accident risk.