Improving PPV prediction in open-pit blasting through Cubist-based feature enrichment and machine learning models

Blast-induced ground vibration poses serious environmental risks in open-pit mines. To improve prediction accuracy, this study introduces a machine learning-based feature enrichment technique using the Cubist algorithm, which generates an additional input variable to enhance the original dataset. Four machine learning models – SVM, RF, k-NN, and GBM – were developed and evaluated on both original and enriched datasets. Results demonstrate that Cubist-based enrichment significantly improves prediction performance, reducing overfitting/underfitting issues. This approach highlights the potential of combining feature enrichment with advanced machine learning models to achieve more accurate and reliable vibration prediction in practical mining applications.