Abstract: Mine pressure prediction is the key and difficult point of coal roof disaster prevention. At present, the popularity of hydraulic support electro-hydraulic control system makes the pressure data can completely and quickly reflect the change of roof pressure. However, due to the large amount of data and many interference information, it is difficult to complete it only by manual analysis. In view of this situation, a method that can automatically extract the cycle cycle, predict and divide the pressure area of the support pressure data is designed. Firstly, according to the working characteristics of the hydraulic support, design rules, and extract the pressure data under normal working conditions as the cycle data. Then, through the clustering method, combined with the pressure data of each support, the significant pressure area is divided. Finally, the LSTM algorithm is used to predict and analyze the cycle pressure data. The feasibility of this method is proved by practice, which provides ideas for the study of mine pressure law and safe mining.