Globally, mining companies attach utmost importance to ore production and cost-effective methods for implementing them, from when ore is excavated from open pits to when it reaches metallurgical plants for processing. Machine learning tools are beginning to take center stage in most systems being implemented in open-pit mines, helping to mitigate challenges associated with data handling and meet desired production targets. The real-time data from mobile mining equipment showing actual performance is stored in databases. The data from these databases is then used to train and test the developed hybrid model. The present work uses a hybrid model XGBOOST-RF with Bayesian optimization (XBOREOPT) to predict the ore production at Kansanshi open pit mine. The hybrid model could predict ore production for the open-pit mine and intelligently present the results. Metrics were used to evaluate the hybrid model’s performance and ensure it was meaningful, effective, relevant, and accurate in making predictions. The results in this study showed a coefficient of determination of 0.9999 for the hybrid model compared to 0.5382 and 0.3729 for linear regression and K Neighbors regressor, respectively, proving that the XBOREOPT model performed better.
Mkokweza et al. (Thu,) studied this question.