Abstract: In order to fully explore the variation rule of gas monitoring data and accurately predict gas concentration in working face, an intelligent prediction model of gas concentration in working face based on CS-LSTM was proposed. The spline interpolation method was used to interpolate the missing values of the monitoring data gas depth, and then the training samples were obtained by dimensionless processing; the Cuckoo Search (CS) algorithm was used to optimize the four hyperparameters of LSTM, including the number of hidden layers, the number of fully connected layers and the number of corresponding neurons, to establish the optimal gas concentration prediction model and predict the gas concentration of the working face in the next 12 h. The results show that, compared with LSTM and LSTM model based on Genetic Algorithm (GA), CS algorithm has better global optimization ability under the same number of iterations, and avoids the disadvantage of GA that is easy to fall into local optimal effectively; the Root Mean Square Error (RMSE) of the prediction model based on CS-LSTM is 0.023, compared with the other two models, this model has higher accuracy and better effect.