Advanced machine learning techniques for predicting dump slope stability in Indian opencast coal mines

The majority of India's coal is mined using opencast methods, which causes more waste dumps to be formed and stability problems. There is a higher chance of dump instability because to the 1148 million cubic meters of overburden that Coal India Limited (CIL) has removed in the past few years. Complex calculations make dump slope stability studies complicated and time-consuming. Analytical and numerical methods are needed to calculate factor of safety (FOS) of dump slope. This research bridges traditional geotechnical methods with emerging computational approaches by integrating advanced ML techniques with rigorous statistical evaluation and a comprehensive dataset to improve dump slope stability prediction accuracy, reliability, and applicability. With so many available options, picking the best ML model can be a challenge. Consequently, for the purpose of this research, the authors selected models using the Lazy predict AutoML algorithm. Using six base models-Gradient Boosting (GBM), Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting (XGB), Histogram Gradient Boosting (HGB), Nu-Support Vector Regressor (NuSVR), Extra Tree Regressor (ETR), with Stacking Ensemble, and H2OAutoML-this study proposes an effective method for analysing dump slope stability. In preparation for model calibration and evaluation, databases of 2250 datasets were created. The output is the SLIDE computed factor of safety, and the inputs are six influential parameters such as cohesion (c), angle of internal friction (ϕ), unit weight (γ), overall bench height (H), natural moisture content (m), and overall slope angle (β). The coefficient of determination (R squared or R²), mean square error (MSE), mean absolute percentage error (MAPE), root mean square error (RMSE), and mean absolute error(MAE) were used for evaluating the performance of all models. The H2O Auto ML performed best model in comparison to other ensemble models. This research also makes use of the Shapley additive explanations (SHAP) technique to determine which of the six inputs is most crucial. This study shows that sophisticated ML approaches improve dump slope stability prediction in Indian opencast coal mines.