<p>Score matrix of the Random Forest models trained with the optimised hyperparameters and a 1:1 ratio between settlements and pseudo-absence points.</p>

The Objective value (Recall) column reports the cross-validated recall on which the hyperparameter selection was based, obtained from the grid search described in the Methods. All remaining metrics (Recall, Precision, Sensitivity, Specificity, F-measure, Accuracy and Cohen’s Kappa) derive from a second, independent model trained with those hyperparameters on 80% of the cases and evaluated on the held-out 20%.</p> <p>(XLSX)</p>