When using long short-term memory (LSTM) models for predicting civil aviation wind shear risks, there are issues such as performance being greatly affected by hyper parameter configuration, and the optimization process being complex and time-consuming. To improve the predictive performance, applicability, and accuracy of civil aviation wind shear risk prediction models, an elite genetic algorithm (eGA) optimized LSTM civil aviation wind shear risk prediction model, namely eGA-LSTM, was proposed. Firstly, the long-term memory capability of LSTM was used to capture the temporal characteristics of wind shear. Secondly, elitist genetic algorithms were introduced to optimize the hyperparameters of the LSTM model, which ensures that excellent characteristics can be preserved and effectively inherited, thereby avoiding the problem of losing excellent individuals that may occur in the iterative process of traditional genetic algorithms, accelerating the hyperparameter optimization process, and improving the generalization ability of LSTM. Historical information on wind shear unsafe events in civil aviation were gathered. Wind shear risk value for 64 months were calculated. Data were divided into training and testing sets for model validation, and compared with a simple LSTM model. The results show that in wind shear risk prediction, the mean square error (MSE) of the eGA LSTM model decreases by 15.49% compared to the simple LSTM model, and there is a significant improvement in accuracy. The model fits the historical sequence of monthly wind shear risk well, demonstrating higher accuracy and reliability in wind shear risk prediction and warning.