Unmanaged gases inside the mine airways are hazards to health and explosions, especially methane (CH4) in coalmines. Temperature rise caused by heat release from the strata and machinery is another factor that may harm the health and safety of workers in underground mines. Control of methane and other gas components as well as high temperature near a working face requires overall and localised ventilation management and adequate mine cooling systems. Continuously monitoring the in situ, atmospheric conditions as well as the amount of contaminant gases, especially methane, are important factors for predicting the necessary actions for keeping the mine a safe and healthy place for workers. Studies are reported for predicting ahead of time methane concentration variations inside underground mines using long-short-term memory (LSTM) artificial recurrent neural network, time-series regression predictor (time series filter), as well as transport model-based methods. Different combinations of the variables and techniques are tested in the LSTM model to find best results for accuracy and applicability. Forward time step variations are tested in time-series regression models to explore the best prediction outcome. The results show that the LSTM and time-series techniques perform similarly and both are sensitive to sliding window sizes and the number of forward-step predictions. The time series filter showed to be extremely faster than the LSTM model and presented a higher accuracy using the first order fitting, especially when the filtered data is used for training and predicting. The transport model used to back-calculate the line source responsible for the increase in methane concentration in different locations inside the mine provides an output that is slightly better than the original data for the time series filter to make the predictions, which gives basically the same relative error compared to the predictions using the original CH4 concentration using both the time series filter and the LSTM Recommendations are discussed for transport model-based solutions for forward predictions combining their parameter identification with those used for NN model training. For the case studied in this paper, the predictions are acceptable, but more tests are need to increase the accuracy and reliability of the predictions from the models and to verify the applicability to the mining industry.