XploreIQ: Successfully Using Machine Learning in Mineral Exploration

Exploration expenditure has significantly increased since 2010, shifting to deeper domains readily available discoveries are progressively exhausted. In the same period, the industry discovery rate has decreased by more than 50 percent, putting forward the question “are we using a fully optimized targeting process?” In 2016, SGS Geological Services team lead by Guy Desharnais, won the Integra Goldrush Challenge using an innovative combination of mineralized vector load in block model, filtering it through machine learning algorithms to produce the next generation of exploration targets. This challenge opens the mind of many explorer about the use of new technologies and how to use them efficiently on any type of deposits.

For the past decade various types of algorithms, including decision trees and stumps boosting enhanced with domain adaptation, were adapted by the mining industry and used on different global projects for targeting purposes. Phylogenetic algorithms were also integrated into the toolbox to answer questions related to geological uncertainties and rock classification using geochemical dataset. Application are multiple in the mining industry from exploration targeting to block model ore to processing reconciliation. This presentation will focus on exploration targeting with three successful case studies for Gold and PGEs using different algorithms and softwares. Limits of the technique and challenges for the future will also be part of the discussion.