Gold Prospectivity Mapping Based on Fractal Information Extraction and Interpretable Machine Learning: A Case Study in Damaoqi, Inner Mongolia, China

　　The Damaoqi area in Inner Mongolia is characterized by a complex metallogenic background where early-stage metamorphic iron deposits are spatially superimposed with late-stage hydrothermal gold and fluorite mineralization. This superposition results in mixed geochemical signals, posing significant challenges for target-specific gold prospectivity mapping. To address this issue, this study proposes an integrated predictive workflow that fuses fractal information extraction with interpretable machine learning. The workflow begins with a centered log-ratio (CLR) transformation of the regional geochemical data to mitigate the closure effect of compositional data. On this basis, Principal Component Analysis (PCA) is used to reveal the complex superposition of the different metallogenic systems, and the Sequential Binary Partitioning (SBP) technique is subsequently applied to extract corresponding element assemblages. Following this, the Spectrum-Area (S-A) fractal model is employed to decompose the key geochemical fields, quantitatively separating the background field, which represents the geological "source, " from the anomaly field, which indicates the "orebody" or trap. Finally, using known gold deposits as positive samples, a predictive model was constructed based on the XGBoost algorithm. The geological interpretability of this "black-box" model was then analyzed using the SHAP (SHapley Additive exPlanations) method.
　　The results indicate that: (1) The combination of PCA, SBP, and the S-A fractal method, applied after CLR transformation, successfully distinguished the geochemical characteristics of the early metamorphic iron system from the later hydrothermal gold-fluorite system. (2) The optimized XGBoost model demonstrated excellent predictive performance, achieving an Area Under the Curve (AUC) value of 0.92. (3) SHAP analysis confirmed that the fractal-decomposed geochemical indicators and fault structures are the most critical factors controlling gold mineralization, and their importance ranking is highly consistent with geological metallogenic patterns. (4) The final mineral prospectivity map successfully delineated two high-potential exploration targets.