Abstract: Rock pressure prediction is an important means of management and disaster prevention for mine roof. It has always been one of the difficult problem in the area of mine roof control. At present, there is lack of mature, reliable dynamic prediction methods. In this paper, the field measurement, data mining and theoretical analysis methods were adopted, and the double cycle analysis and prediction model for rock pressure was constructed based on the periodic variation law of support load in time and space. On the basis of the research on the interaction relationship between the surrounding rock and the support, three basic resistance increasing functions were obtained, including index functions, linear functions and logarithmic functions. Through the analysis of the measured data, it was considered that the support load time series curve of No.12401 working face is mainly divided into four types, including index functions, logarithmic functions, logarithmic and linear functions, and logarithmic and index functions. According to the sliding window dynamic prediction method and the maximum fitting based on the principle of goodness, a regression prediction model of the working resistance of the support in the coal mining cycle was constructed. The two methods of systematic clustering and K means clustering were used to classify the end of cycle resistance to construct the template curve, based on extension window prediction method and maximum matching criterion. The prediction model through matching was constructed. Through support resistance in No.12401 working face and software development, the double cycle prediction method of support resistance was established, which realized the collection, preprocessing, analysis of rock pressure characteristic indicators, regression prediction, and matching prediction, based on the resistance in the electro hydraulic control system. The prediction accuracy of support load and end of circulation resistance is over 86%.The double cycle prediction model can provide reference for the pre development of surrounding rock control strategy.