Abstract: Aiming at the problems of large discreteness of early warning results and prediction positions of rock burst, a multi-source information fusion early warning technology is studied, and a multi-source information fusion depth prediction model based on machine learning algorithm is established. Through the analysis of rock burst events, the main causes, characteristics and influencing factors of rock burst accidents are analyzed. Taking the compressive strength, tensile strength, elastic energy and ground stress of coal and rock as the main indicators of rock burst prediction, the depth neural grid prediction model is established to determine the weight of different prediction indicators. Through big data analysis and limited data, the occurrence of rock burst is predicted. Taking the occurrence of rock burst is predicted. Taking a coal mine in Shaanxi Province as an example, the depth neural network model is verified, and the effectiveness and correctness of the prediction model of rock burst are verified through field measurement.