中国矿山工程

The XGBoost ensemble learning algorithm has excellent performance in solving complex nonlinear relationship problems. In order to accurately predict the surface subsidence caused by mining, the genetic algorithm was introduced to optimize the XGBoost model, and the GA-XGBoost combined model was developed using the python programming language. Firstly, the hyperparameter vector of XGBoost is randomly initialized, the prediction error of the model is obtained after training and testing, and the XGBoost model is optimized by GA, and finally the XGBoost model with the best performance is obtained, and 78 domestic coal mining subsidence data are predicted. The prediction results show that the R2 (coefficient of determination) of the prediction results of the GA-XGBoost model is 0.9318, the RMSE (root mean square error) is 0.3989, and the MAE (mean absolute error) is 0.2989. Compared with ensemble learning models such as single XGBoost, random deep forest, and Gradient Boost, the GA-XGBoost model has higher mining subsidence prediction accuracy.