Abstract: To enhance the accuracy of airflow prediction in mine ventilation systems, this study proposes a BP neural network prediction model optimized by an improved Sparrow Search Algorithm (SSA), aiming to address the issues of random initialization and slow convergence in traditional BP neural networks. The improved SSA integrates Cauchy mutation and reverse learning mechanisms, thereby strengthening its global search capability and accelerating convergence. By using the improved SSA to optimize the initial weights and thresholds of the BP neural network, an enhanced SSA-BP prediction model is constructed. Mine ventilation volume is affected by multiple factors and exhibits strong nonlinear characteristics. In this study, five models—LSTM, BP, PSO-BP, SSA-BP, and the improved SSA-BP—are developed and comparatively evaluated. Model performance is assessed using Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE), Root Mean Square Error (RMSE), and the fitting coefficient (EC). The improved SSA-BP model achieves an EC of 0.98, which is 0.10, 0.06, 0.05, and 0.03 higher than those of the LSTM, BP, PSO-BP, and SSA-BP models, respectively, and it also outperforms all comparison models in terms of MAE, MAPE, MSE, and RMSE. These results indicate that the improved SSA-BP neural network model can significantly enhance airflow prediction accuracy and provide reliable data support and predictive reference for the development of intelligent mine ventilation systems.