Stacked Generalization for Improved Prediction of Ground Vibration from Blasting in Open‑Pit Mine Operations

Blasting generates undesirable ground vibration with adverse effects on humans, the environment, and structures. Numerous studies have used various techniques including machine learning (ML) to predict the peak particle velocity (PPV) caused by blasting in surface mines. These studies used the validation-set approach to evaluate the performance of their predictive models. In this study, Random Forest (RF), Gaussian Process (GP), and Gradient Boosting Machine (GBM) were applied for the prediction of PPV using data from a large open-pit gold mine in Ghana. An ensemble model based on a stacked generalization approach was then built on these three base models to significantly improve the PPV prediction performance. The model evaluation was done using repeated tenfold cross-validation which resolves the challenges with the validationset approach for small datasets. The dataset used in this study comprised a total of 196 measurements taken from different blasting operations. R-squared ( R2 ), mean absolute error (MAE), and root mean square error (RMSE) were used to evaluate the model performance. The results obtained from these models revealed that all the base models did not significantly outperform each other. However, the stacked generalization approach had a better performance, with RMSE = 0.12, R2 = 0.96, and MAE = 0.07. This approach presented an improvement over the existing methods and will be useful for highly accurate predictions of ground vibration during blast design. The application of this model will be useful for efficient and effective blast design.