Prediction of Aviation Safety Using the Combination Model Based on Improved Immune Algorithm

Flight accidents can cause huge casualties and property damage, and establishing an aviation safety prediction model is of great significance for the design and prevention of aircraft operation safety. Aviation safety incidents have the characteristics of multiple influencing factors and randomness. Existing prediction methods are mainly based on time series, with low explicit expression of causal factors, poor interpretability of models, and weak engineering practicality. This article proposes a method for constructing multiple regression analysis(RA) model functions based on unit regression using the “cumulative addition by cause” method and the “cumulative multiplication by cause” method, which improves the modeling efficiency of the multiple regression model; An improved immune optimization algorithm (IA) has been proposed, which enhances population diversity and global optimization capability; Based on this, this article proposes a “Multiple Regression+Neural Network” (IARA+BPNN) safety combination prediction algorithm based on an improved immune algorithm. The prediction target is decomposed into “rule term+correction term”, and the rule term is made explicit through multiple regression model prediction. The global optimization ability of the multiple regression model weight is improved through the improved immune algorithm, and the correction term is predicted through the BP neural network (BPNN) model to enhance the prediction accuracy. Taking the sample data of A-level military support mishaps in the United States as an example, the prediction results show that the “cumulative addition by cause” and “cumulative multiplication by cause” improve the efficiency of constructing multiple regression functions, and the improved immune algorithm improves the prediction accuracy of regular terms. The “IARA+BPNN” prediction model combines the advantages of multiple regression models and BP neural network models, which not only has certain explicit requirements, but also has high prediction accuracy, improving engineering practicality.