Abstract
<title>Abstract</title> <p>The growing demand for renewable energy technologies, Electric Vehicles (Evs), and advanced electronics underscores the vital importance of rare earth elements (REEs) as critical minerals and its associated demand for effective, data-informed exploration strategies. This study aimed at developing an explainable stacked ensemble machine learning framework for the REE prospectivity mapping in Ruri Carbonatite Complex, southwest Kenya, which is underexplored. Geological, geochemical, radiometric and spatial datasets were combined to train a stacked ensemble model with Support Vector Machine (SVM) and Random Forest (RF) as base learners and Extreme Gradient Boosting (XGBoost) as a meta-learner. The performance of the model was tested by 5-fold stratified cross-validation and various classification measures such as accuracy, precision, recall, F1-score, Matthews Correlation Coefficient, Cohen's Kappa, ROC-AUC and PR-AUC. The highest performance was obtained by the ensemble model which had an accuracy of 95.6% and a ROC-AUC of 0.97. The most influential predictors identified by SHapley Additive exPlanations (SHAP) were: Total Rare Earth Oxides (TREO), LREE/HREE fractionation, thorium, uranium, radioactive enrichment indices and structural density. The map of prospectivity, produced using the explainable ensemble learning, identified several REE targets related to carbonatite intrusions, fenitized host rocks, and main structural corridors, highlighting the importance of explainable ensemble learning for REE exploration with reliable and interpretable outputs.</p>