The partnership between mining consulting services provider The MSA Group and Canadian mining software developer Koan Analytics has resulted in machine learning being available to minerals explorers in South Africa and throughout the world.
“The partnership offers access to the most advanced analytics platform geared towards mineral prospectivity mapping and target generation,” says MSA exploration services manager Craig Blane.
He says any company looking to embark on mineral exploration programmes would typically start with an exploration target generation process.
The objective is to use all available data to identify areas where the likelihood of a deposit discovery is high. The data is collated at regional and then deposit scale.
Machine learning technology is not widely used in the domestic exploration space thus far, with only specialist consultancies offering advanced machine learning-driven target generation and prospectivity mapping services.
The partnership, formed in 2020, entails Koan contributing its machine learning and analytics capability, while MSA provides its expert geological input using its expertise on deposit models and mineral systems modelling to interrogate and interpret data to generate exploration targets.
Blane explains that the process starts with building an integrated data repository by ingesting text, table and image-based information from a variety of geoscience datasets, including reports, log sheets, geophysics and geochemistry datasets.
The volume of data that can be fed into a target generation study with the aid of a machine learning analytics platform is substantially greater than that of traditional methods, he adds.
“This allows for new and previously missed targets to be identified and will improve prioritisation and target selection. Further, the time required to generate, and rank targets can be greatly reduced,” he enthuses.
Koan has developed an interface that enables users to operate the platform and identify features of interest using the clusters, graphs and map tools to interrogate the underlying data to develop insights into potential locations for mineralisation.
“The process can be easily adapted to almost any region in the world,” he says.
New Kid on the Block
“As a consequence of machine learning advances in data repository development, new integrated approaches to exploration analytics are emerging,” says Koan Analytics CEO Rob Wood.
“Each piece of data from massive data sources is extracted, labelled and classified, so each individual data artifact or piece of information can be correlated and associated with every other data artifact in the data repository.”
During this process, the data repository develops from merely storing data to holding the innate (semantic) relationships and insights across all geospatially associated information.
Koan’s correlation classification system is persistent – once the system learns something it retains the knowledge and, as new data is added, this information is added to the system’s understanding.
This enables geologists to perform complex analyses across different data attributes such as: lithology, geochronology, structure, deposit types, geophysics, geochemistry, mineral occurrences and mineral alterations.
Machine learning-based analytics gives geologists the capability to investigate the correlations and associations across multiple compound variables.
“Virtually any form of relationship across all data variables can be analysed, opening up whole new fields of exploration analysis,” Wood enthuses.
Wood concludes that Koan is the “only company globally in the resource sector with a universal data repository and with the proprietary machine learning models and analytic processes necessary to develop it”.