The prevention of rock burst is a major problem in coal mining.In recent years,with the gradual shift from shallow layer to deep layer of coal mining,the number of rock burst in coal mines has increased.Rock burst is a serious threat to the safety of coal mine workers and will bring huge economic losses,so it is particularly important to study the prediction of rock burst.The traditional prediction method can only analyze a small amount of precursor information before the occurrence of rock burst,and can not predict the change trend of future rock burst related signals according to the historical information.In order to explore the prediction method of rock burst,the research group selected rocks from coal mine with rock burst,and used TYJ-500KN microcomputer controlled electro-hydraulic servo rock shear rheological test system and SH-II acoustic emission system to carry out rock burst similarity simulation experiment.The compressive strength signals and acoustic emission signals collected in the experiment are fused,and the data are predicted by Long Short-Term Memory neural network(LSTM) with memory properties.The results show that the curve fitting between the predicted data and the actual analysis data is high,and the maximum root mean square error of the data is less than 0.6.The LSTM model has an excellent research prospect for the prediction of rock burst.