Development of a machine learning model to classify mineral deposits using sphalerite chemistry and mineral assemblages

Sphalerite is a commonly occurring mineral in natural systems and a prominent indicator mineral used in resource exploration. Its chemistry and associated mineral assemblages are both controlled by the physicochemical conditions of the local environment, such as temperature and sulfur fugacity. Accordingly, the chemistry of sphalerite and the nature of the associated minerals provide valuable clues about the classification of mineral deposits and their environment of formation. Nevertheless, exclusive reliance on the trace-element chemistry of sphalerite for deposit classification has its limitations given the multitude of factors (For example, deposit type, temperature, pressure, and background concentrations of elements) that affect its chemistry. To address this challenge, we develop machine learning models using the SHapley Additive exPlanations (SHAP) method to assess the importance of sphalerite trace-element chemistry and mineral assemblage information to distinguishing between five petrogenetically distinct mineral deposit types. This contribution demonstrates that a composite model that incorporates both data types markedly improve the accuracy of deposit type discrimination. The composite model is composed of two sub-models, Random Forest (RF) and Extra Random Trees (ERT), which are specifically employed for processing trace-element data and mineral assemblage data, respectively, due to their superior performance in handling these two distinct types of datasets. To simplify the end-user experience, we provide an executable file of the machine learning-based classifier, allowing it to be readily applied as an exploration tool using simple inputs. In summary, this work substantially enhances the confidence of Pb-Zn deposit type classification using sphalerite and introduces an innovative perspective on the application of machine learning to resolving complex geological problems.