Abstract
The ore material extracted from a mineral deposit can have varying geochemical properties, causing processing challenges. Nevertheless, the orebody can be divided into homogeneous groups with similar compositional footprints. Yet, current grouping approaches are subjective and require a significant human workload due to the lack of an optimized and objective methodology. This research developed a new data-driven workflow using a PaCMAP dimension reduction and a novel clustering algorithm, Clusterdc, to automatically identify compositional clusters in a geochemical dataset, reflecting the orebody compositional variability. The research project outcomes displayed that the proposed approach could assist geoscientists in identifying compositional clusters, which is a critical initial step in guiding sample selection for further metallurgical testworks in order to enhance mining companies’ orebody knowledge.