APPLICATION OF THE MINERAL SYSTEMS AND ARTIFICIAL INTELLIGENCE APPROACH TO THE PROSPECTIVITY OF PORPHYRY-SKARN DEPOSITS IN THE ANDAHUAYLAS-YAURI BATHOLITH *K.R. Balbuena1, G.M. Baquerizo1, L.B. Espinoza1, J.A. Milla1, I.A. Montalván1 & J.A. Torres1 1Department of Geological Engineering, National University of San Marcos, Lima, Perú (*Presenting author: kevin.balbuena@unmsm.edu.pe) ABSTRACT Regional targeting of porphyry–skarn systems in southeastern Peru remains challenged by the subjective integration of multiple geological and structural proxies. This study presents a regionalscale decision-support framework that combines the Mineral Systems approach with supervised machine learning to prioritize prospective domains within the Andahuaylas–Yauri Batholith metallogenic province. The methodology is structured around three critical elements of the Mineral System concept—Fertility, Geodynamics, and Architecture—following the process-based framework of McCuaig & Hronsky. Publicly available geological and structural datasets were translated into spatial proxies representing permissive conditions for porphyry–skarn mineralization at regional scale. These predictors were integrated using Random Forest and Artificial Neural Networks to assess their relative influence on the spatial distribution of known deposits. Both algorithms consistently identify Fertility—primarily expressed by the distribution and characteristics of Eocene–Oligocene magmatism within the batholith—as the dominant permissive factor controlling regional prospectivity. Architectural elements, including major translithospheric and transverse fault systems, exert a secondary yet spatially significant influence. Geodynamic parameters were incorporated in a simplified manner due to limitations in regionalscale datasets and therefore represent broader tectonic context rather than direct predictive variables. The framework is explicitly scale-dependent: it delineates permissive regional domains rather than predicting deposit-scale localization or threshold-controlled trapping processes. Prospectivity outputs reproduce the principal known porphyry–skarn districts and highlight additional favorable sectors outside historically explored corridors, suggesting areas for further geological evaluation. By embedding process-based geological reasoning within transparent machine learning workflows, the proposed approach reduces subjectivity in early-stage targeting while maintaining conceptual consistency with the Mineral Systems paradigm. The methodology offers a reproducible and scalable tool for regional exploration prioritization in porphyry–skarn provinces, while acknowledging the constraints imposed by data resolution, proxy selection, and scale. KEYWORDS Mineral Systems Framework; Random Forest and Neural Networks; Magmatic Fertility; Structural Architecture; Regional-Scale Targeting
RkJQdWJsaXNoZXIy MTM0Mzk2