Leg pressure data from all shields of a longwall face are monitored and recorded in the surface computer. An algorithm is developed to detect peak pressures or periodic roof weightings from these pressure data. The intensities and locations of periodic roof weighting are further analyzed using artificial neural network for forecasting of forthcoming shield pressures. The network was trained using data 153 m (500 ft) of face advance. Shield pressures are forecasted for the successive nine mining cycles or approximately 9 m of face advancement. The results obtained validate the efficacy of the developed model.