Abstract: Aiming at the engineering problems of traditional mine roof pressure monitoring technology, such as insufficient warning reliability, limited prediction accuracy and low level of intelligence, we put forward the mine pressure intelligent early warning method system integrating machine learning, constructed the mine pressure anomaly identification model based on the time-sequence load characteristics of hydraulic bracket and the cycle pressure span through integrating the multivariate regression analysis and the hierarchical clustering algorithm, and developed the mine pressure intelligent early warning system with the ability of autonomous decision-making. It has developed an intelligent early warning system for roof disaster with autonomous decision-making capability. The industrial test data show that the system's functional completeness index reaches 0.93, and achieves 92.6% accuracy in cycle pressure prediction in Dayangquan coal mine, which significantly improves the intelligent level of mine pressure disaster prevention and control.