Uso de machine learning para validar las litologías de la zona Alamala-Catamayo

Abstract

Abstract: This study was based on the geochemical results collected in the Alamala Catamayo area by Geology and Mines students in 2018. The database includes 64 sampled points with lithological and geochemical information, using a 1000m x 1000m and 1000m x 1500m mesh. Geostatistical and machine learning techniques were applied to generate geochemical maps of various elements Al2O3, SiO2, CaO, Fe2O3, K2O, S, MnO, TiO2, MgO; and a predictive map of lithologies. The analysis includes tools such as histograms and correlations, achieving the classification of quantitative and categorical variables through geostatistical models. Variographic analysis was used to understand geochemical dispersion, and supervised classification, especially using decision trees, allowed lithologies to be determined in unsampled areas. The results revealed a deeper understanding of the areal composition and significant improvements in the predictive ability of lithologies, supporting the comprehensiveness of the study approach.