In the study of deposits and the exploitation of mineral resources, traditionally, the decisions of the geographical positioning of new drilling wells necessary for the generation of geomodels are made subjectively, under a professional criterion that often responds to the experience and assumptions of the geologist.and not to a methodology based on the scientific method.Therefore, there is a need to create a non-subjective decision-making methodology in mineral resource exploration, with greater reproducibility and less uncertainty.Characterizing a deposit begins many times with field work that defines the taking of samples and the measurement of its physical and geochemical parameters, and it must end with the generation of geologically plausible models that favor decision-making on the early stages of development of a mining project.Obtaining these models is a great challenge for Colombian companies and much more because they do not have comprehensive methodologies developed thinking about the specific conditions of the place of study, and supported by computational tools for the treatment and analysis of the data that result from field work.and in the laboratory for collected samples.This work presents the first stage of the development of a methodology for the generation of mineral deposit models using machine learning methods (Machine Learning, ML).