A Construction Safety Accident Prediction Model Based on Genetic Algorithm Optimized Long Short‐Term Memory Neural Network

Construction safety accidents remain a major threat to life and property in the construction industry. To improve accident early‐warning capability, this study develops a prediction model based on a genetic algorithm‐optimized long short‐term memory neural network (GA‐LSTM). Using statistical data on production safety accidents in China’s housing and municipal engineering sectors from 2015 to 2020, the proposed GA‐LSTM model is trained and validated, and its performance is compared against three benchmark models: the back propagation neural network (BPNN), the genetic algorithm–optimized BPNN (GA–BPNN), and the LSTM model. The results show that GA‐LSTM achieves the highest prediction accuracy, the mean absolute error (MAE) of GA‐LSTM was reduced by 56.83%, 45.35%, and 36.09%, respectively, while the mean absolute percentage error (MAPE) was reduced by 60.77%, 42.73%, and 40.0%, respectively. Key findings reveal that GA effectively enhances model optimization by rapidly identifying optimal hyperparameter configurations in LSTM, thereby improving both convergence efficiency and local search capability. The research outcomes provide actionable insights for implementing intelligent safety monitoring systems and establish a methodological foundation for applying hybrid AI technologies in infrastructure risk prevention.