Abstract: In order to overcome the difficulty in predicting the manifestation law of roof mining pressure on fully mechanized working face, based on the big data of support resistance, a multivariate linear area prediction model was designed by machine learning language about three important parameters, namely the starting position of the press, the press strength of roof and the interval of the press. This paper introduces the establishment process of the mathematical model of the machine learning prediction algorithm based on big data and the test of the prediction model, taking the original data of 30 groups of mine pressure monitoring as an example, the mathematical model of big data analysis was constructed. The results show that the model is feasible and the prediction algorithm is determined.