Abstract:In order to improve the accuracy of the prediction results caused by underground mining, we propose a surface subsidence prediction model of BP-Adaboost, which fuses chaos residuals. Taking the measured value of 1312 (1) of Gubei mine as an example, we use the BP-Adaboost models, the BP neural network model, and BP-Adaboost model fused with chaotic residuals to make one-step and multi-step predictions for the stability and active period of the maximum sinking value point, respectively. The experimental results show that BP-Adaboost model fused with chaotic residuals has the highest accuracy in both one-step prediction and multi-step prediction, especially for one-step prediction.