155 ● Geochemistry: Point data from sediments (light-blue-circles) and rocks (orange-square) were processed by a Chemical Association Index (Balaram, 2022; Holland, 2008; Chen, 2015), summing normalized Z-scores As-Cu-Mo-Pb-Sb-Se-Zn, consistent with multivariate geochemical analysis principles (Barnett, 2017; Taherdoost et al., 2014). ● Geophysics: The RTP magnetic grid was reclassified using natural breaks classification to preserve magnetic contrasts while avoiding over-parameterization, ensuring that magnetic variability associated with intrusive centers retained without exceeding the resolution supported. 2.2. Heuristic Methods of Integration These heuristic methods were implemented as benchmark baselines to compare traditional expertdriven approaches against data-driven ML models. 2.2.1 Simple Average: For this method, the average value was calculated from the layers of faults, geology, geophysics, geochemistry, and the 46 resulting combinations from ASTER imagery (Figure 2). 2.2.2 Empirical Weighted Mean: This method applied a knowledge-based weighted linear combination. The following weights were assigned: Faults and Geology (30), Sediment Geochemistry (28), Rock Geochemistry (24), and Geophysics (20). ASTER alteration layers received in the porphyry alteration model (advanced argillic: 10, alunite: 10, sericite-illite: 9, jarosite: 9, propylitic: 8, kaolinite: 7, chlorite: 7, phyllic: 7, illite-kaolinite-sericite: 6, epidote: 6, hematite: 6, iron oxides: 4, carbonates: 3, biotite: 3, quartz: 1). Weights were assigned between 6 and 10, according to their importance (Figure 2). Figure 2 – Simple Average and Empirical Weighted Mean Heuristic Methods 2.3. Supervised Machine Learning Algorithms All ML models followed a common workflow: imputation of missing values (primarily using the median), stratified splitting of the data into 70% for training and 30% for validation and using the location of known deposits ("yac") as the target variable for prediction. Grid cells intersecting known deposits were labeled as positive (“yac” = 1), and remaining cells as negative (0). Although the response variable was binary, models generated continuous prospectivity scores, enabling evaluation with regression metrics (R², MAE, RMSE) and classification metrics (AUC, Precision, Recall, F1) using the threshold on the validation dataset. The selected algorithms represent diverse modeling paradigms, including ensemble tree methods, kernel-based methods, distance-based classification, and semi-parametric regression models. Algorithms, key hyperparameters, and performance metrics are detailed below:
RkJQdWJsaXNoZXIy MTM0Mzk2