A Review of Machine Learning-Based Geochemical Signature Analysis for Mineral Prospectivity Mapping.

Abstract

Abstract: Geochemical signature analysis has been a basic technique of mineral exploration over the years, but the nonlinear and complicated nature of multi-element geochemical data has proven hard to capture using traditional tools of statistical analysis. This is because the incorporation of machine learning algorithms into geochemical analysis is a paradigm shift that will allow more sophisticated pattern recognition and predictive modeling of mineral prospectivity maps. This review summarizes the existing information on machine learning as applied to the geochemical signature analysis, including the theoretical basis of the method, algorithms, and application in different geological environments. We delve into how supervised approaches to learning, including Random Forest, Support Vector Machines, and neural networks have revolutionized the field of anomaly detection and target generation and unsupervised approaches to learning, including clustering algorithms and dimensionality reduction procedures, are used to discover the unknown geochemical worlds. A review is done of the successful case studies using various types of deposits and in geological environments with a focus on uses in underexplored areas such as African metallogenic provinces. The problematic issues, such as the complexity of data preprocessing, the interpretability of the models, and the ability to generalize and apply the models to various geological settings are addressed. New directions in architecture, like deep learning and explainable artificial intelligence, as well as multi-source data integration, are also indicative of more advanced exploration processes. This detailed discussion shows that geochemical analysis based on machine learning does not only increases the level of target identification but also redefines the principles of exploration, providing avenues to exploration in both developed and frontier geology and responding to the pressing demand of new mineral resources in an era of energy transition.