Abstract: In order to predict the overlying rock movement during the mining, based on the orthogonal experiment design, the numerical simulation program FLAC^3D was used to analyze the overlying rock movement. The numerical simulation program was composed of 16 kinds of combinations which was determined by four factors of mining thickness, mining depth, mining width, properties of overlying rock. According to the results of the numerical simulation program FLAC^3D,the overlying rock movement based on the BP neural networks was built and the learning samples of neural network was determined. The result of BP neural network was compared with the result of the numerical simulation program. The results shown that the results obtained by BP neural networks were close to reality, the error was low, which has a high credibility.