Featured Articles

“Recent advances in cloud computing may be allowing geostatistics to crunch more data, faster, but its merging with machine learning techniques that will continue to be the most exciting development in the next decade. This opens an entirely new landscape for applications in the mining sector.

In simplified terms, geostatistics are a set of model-driven algorithms, and machine learning can be seen as a data-driven approach. Combining them to get the best of both worlds is the holy grail of geological modelling. This is what we are doing at DataCloud, especially when a client asks us to integrate massive datasets and visualise their orebody in new ways to gain new insights.

A mine site is full of orebody knowledge, evolution of rock, geochemistry, mineralogy, hyperspectral, geometallurgy and more. All this data collected is challenging to incorporate into a consistent spatial model: geostatistics can struggle with large amounts of disparate data types. On the other hand, it can be difficult to incorporate spatial constraints in machine learning models. Now they can support each other: model-driven geological features incorporated into machine learning frameworks that explicitly consider spatial correlation as well as uncertainty. There is no silver bullet here. Geostatistics and machine learning are powerful tools on their own, but building a framework that takes advantage of the strength of both approaches is the best way forward to invent powerful new algorithms and workflows.”