based on block-level placement and the reconstructed heap-state response, not direct per-block measurements. All quantitative values are withheld for confidentiality. Figure 3 –Predicted Au vs Actual Au Scatter Plot using linear regression with engineered features model. During the training period, the model explained approximately 64% of the variance in monthly gold production (R² = 0.64). When evaluated on unseen validation and testing periods, the model maintained predictive capability, achieving R² values of 0.38 and 0.37 respectively. The relatively moderate decrease from training to validation indicates limited overfitting and acceptable generalization capacity. These results suggest that linear combinations of engineered geometallurgical and operational features capture a meaningful portion of the production dynamics. However, the unexplained variance highlights the presence of nonlinear behavior, long-memory effects, and dynamic inventory interactions inherent to heap leach systems. The model therefore provides a stable reference benchmark for evaluating more advanced dynamic or physics-informed machine learning approaches. Figure 4 – Spatial Maps of heap leaching performance. A spatial Heap Leach Performance Map was developed to evaluate the distribution of recovery efficiency across the leach pad. The map integrates weighted heap height and cumulative soluble gold placement, allowing spatial identification of zones exhibiting relatively higher recovery response within the analyzed time window. 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Predicted Au Actual Au
RkJQdWJsaXNoZXIy MTM0Mzk2