Exeter DPhil Student Applies ML to Advance Mineral Exploration

Felix Sihombing (2021, DPhil Earth Sciences), a Jardine Scholar at Exeter College, is pioneering the use of graph deep learning to enhance mineral exploration techniques. With a background in field geology and academic research, Felix is developing innovative tools that integrate graph deep learning and unsupervised algorithms to better identify mineral deposits. 

Before commencing his doctoral studies at Oxford, Felix worked as an exploration geologist in Indonesia, including roles at Freeport-McMoRan’s subsidiary and Micromine Indonesia, a geological software company. He later served as a lecturer in structural geology and ore deposit studies at Universitas Indonesia, where he remains affiliated. His diverse experience positions him uniquely to bridge the gap between field geoscientists and data scientists. 

In 2024, Felix co-authored a study published in Ore Geology Reviews that explores how graph neural networks (GNNs) can enhance mineral prospectivity mapping (MPM). The research highlights a major limitation of conventional machine learning techniques, which often fail to account for spatial relationships in geological data by treating data points as isolated entities. By redesigning exploration datasets into graph formats that emphasise spatial and relational context, Felix and his co-authors demonstrated that GNNs can significantly outperform traditional methods, particularly when working with imbalanced datasets containing more barren than mineralised data points. Applied to copper, iron, and tin deposits in the southern UK, their approach yielded improvements across key metrics, suggesting that GNNs could offer a powerful tool for mineral exploration in areas with incomplete geological records.  

To learn more about Felix’s work, visit his profile on the Department of Earth Sciences website.