Deformation analysis and prediction of an instability area in an open pit slope in Canada using multi-source field data and machine learning algorithms

Slope deformation is a critical indicator of pit wall stability in open-pit mining, yet its prediction remains challenging due to the combined effects of mining operations and environmental conditions. This study investigates the use of machine learning to predict deformation trends in a large open-pit mine in British Columbia, using multi-source field-monitoring data. Slope displacement was continuously measured using ground-based radar, and associated datasets, including bench excavation records, blasting vibration, precipitation, pore water pressure, and temperature, were integrated into the analysis. Deformation analysis indicates that bench excavations near fault-controlled unstable zones significantly accelerate slope movement, whereas blasting-induced vibrations have a limited direct influence. Three machine learning algorithms, linear regression, regression trees, and support vector machines, were evaluated under both random and chronological data-splitting strategies. While all models reproduced overall deformation trends under random data splits, regression tree and support vector machine models demonstrated superior performance in forward prediction scenarios. Short-term forward predictions accurately captured deformation velocities and peak movement periods, with regression trees providing the best predictive accuracy. The results demonstrate that machine learning, when trained on comprehensive field datasets, can provide reliable short-term forecasts of slope deformation, thereby supporting operational decision-making and risk mitigation in open-pit mining.