Abstract: The air temperature in the intake shaft of high-temperature mines is affected by multiple factors, and there are complex and dynamic nonlinear relationships between these factors, which requires the air temperature prediction model to have the ability to update and adapt to new data sources and conditions in real time. However, this dynamic change increases the difficulty of model learning and training, which in turn affects the accuracy of prediction results. To address this issue, a high-temperature mine air temperature prediction method based on L-M neural network is proposed. Using DEMATEL method to screen and determine these complex and dynamic influencing factors, to ensure that the selected indicators can accurately reflect the impact of mine environment on air temperature. Build a wellbore air temperature prediction model based on the selected input indicators. To further enhance the learning and fitting ability of the model, the L-M algorithm is applied to optimize the neural network. The experimental results show that the maximum prediction error of this prediction method does not exceed 2 ℃, and the fitting coefficient is stable at around 0.95, fully demonstrating the accuracy and reliability of this method in predicting the air temperature of the intake shaft in high-temperature mines. Compared with other traditional prediction methods, this method not only significantly improves prediction accuracy, but also provides more reliable and scientific decision-making basis for mine ventilation management. Therefore, the L-M neural network-based method for predicting the air temperature in high-temperature mine intake shafts provides an effective and practical means for achieving accurate air temperature prediction.