Geoscience Australia is training computers to find hidden resources.

The federal government agency is developing developed machine learning algorithms to hunt for minerals beneath the Earth's surface.

The experts have been working on the Uncover ML project for several years, finding ways to predict the thickness of the Earth's cover and what minerals are hidden below.

The project automates the process of poring through datasets one-by-one, using regression analysis to process magnetic data, satellite data, geochemistry data and other data to form a single prediction model.

Geoscience Australia has been testing its methods to see if they are more efficient at finding gold and uranium deposits than tedious processes that rely on human analysis.

The experts estimate that the machine learning model is already able to match the effectiveness of manual processes about 80-90 per cent of the time, but much faster.

The machine learning applications are being run on the NCI supercomputer in Canberra.

Geoscience Australia has already released Digital Earth Australia, an analysis platform that can peer through cloud cover on satellite imagery, clearing up the data for scientific analysis by humans.

Digital Earth can also observe water from space, analysing the location of rivers and other water corridors over the past 10 years to predict water presence during floods and other changes.

They are now working on improving the machine learning initiative to predict the presence of groundwater.