Risk Prediction of a Domestic Regional Transportation Aircraft Based on Composite Models

To study the aviation operational risk status of a certain type of domestic regional transport aircraft, a risk prediction index system is established by considering factors such as event level, event type, event quantity, event stage, and event cause. Based on the unsafe event data of aircraft in the past 6 years, the Time Series Analysis and Prediction Model (ARIMA) is applied to reflect the seasonal characteristics of risk and fit the linear relationship of risk data. The BP neural network is used to reflect the sudden characteristics of risk and fit the nonlinear relationship of risk data. The risk prediction results calculated for each model are weighted using CRITIC objective weighting method to output the combined prediction results. The results show that the combination of ARIMA and BP neural network models fits the historical sequence of risks well, and the prediction accuracy is acceptable. This study can provide data support for analyzing the safety and operational trends of airlines, and also provide assistance for airlines to develop targeted risk prevention and control measures.