Random Forest–Based Coal Mine Roof Displacement Prediction and Application

Abstract

Coal mine roof accident is one of the most important geological disasters faced by coal mine, and roof displacement is an important index to measure the effect of roadway control and construction safety. Therefore, this study puts forward the machine learning method to study the advance prediction of coal roadway roof displacement. By identifying six important influencing indexes of coal roadway roof displacement, the prediction dataset of coal roadway roof displacement is established and the correlation and importance of the indexes are analyzed. Based on Random Forest (RF), XGBoost, and Gradient Boosting Decision Tree (GBDT) algorithms, three kinds of roof displacement prediction models are established respectively. R2, mean absolute error (MAE), mean square error (MSE), mean absolute percentage error (MAPE), and root mean square error (RMSE) are selected to evaluate the performance of the models. The results show that the RF model has the best performance while the XGBoost model has the worst performance. When RF model is applied to coal mine roadway, the average R2 is 0.92, and the relative error of prediction is 1.76%–9.11%, which indicates that RF model has greater accuracy and applicability in advance prediction of coal roadway roof displacement. It can be used to predict the roof accidents of coal mines to ensure the safety and stability of the roadway enclosure rock during the construction period. The study will be helpful in planning of mining, improving the recovery rate of resources, and promoting the intelligent development of coal mine.