---

Integration of Machine Learning with In-Situ LA-ICP-MS Trace Element Analysis: Multi-Classification Approach Reveals the Hydrothermal Origin of Sphalerite in the Zawar Zn-Pb Deposit, Rajasthan, India.

Gorai S, Dalai B, Kumar TV, Sreenivas B.

JOURNAL OF THE GEOLOGICAL SOCIETY OF INDIA

https://doi.org/10.17491/jgsi/2025/174229 

Geostatistics plays a vital role in mineral exploration by modelling spatial variability of ore grades, optimising sampling, and improving resource estimation with quantified uncertainty. In this study, the authors have integrated geostatistical principles with machine learning algorithms to classify the Zawar Pb–Zn deposit as a Mississippi Valley-Type (MVT) deposit, based on trace element data of sphalerite analysed at CSIR–NGRI. They framed this as a multi-classification problem, using Extremely Randomised Trees (Extra Trees) and Extreme Gradient Boosting (XGBoost) algorithms to process complex, non-linear datasets. This brownfield research approach provides a robust framework that can be extended to greenfield exploration for characterising new mineral deposits.

Fig: t-SNE visualisation plots of mineral deposit classification: Embeddings of training and prediction data for various deposit types (Epithermal, MVT, SEDEX, Skarn, VMS), highlighting the clustering patterns based on elemental features.

---