---

Machine learning applications in trace element analysis: decoding the origin of galena in the Zawar Zn-Pb Deposit, India.

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

EARTH SCIENCE INFORMATICS

http://dx.doi.org/10.1007/s12145-025-01961-3 

This study uses machine learning (Random Forest and Gradient Boosting) on in-situ LA-ICP-MS trace element data (Se, Ag, Cd, Sn, Sb, Tl, Bi) to classify the genesis of galena in the Zawar Zn-Pb deposit, India. Trained on global samples, the models achieved ~97.5% and ~96.4% accuracy, identifying Zawar as the MVT-type (Mississippi Valley type) deposit. t-SNE visualizations shows the distribution of training samples and the 37 predicted samples from the Zawar deposit across the six deposit types using the RF and GB models. The findings support hydrothermal fluid activity during Paleoproterozoic fore-arc spreading and demonstrate ML's potential in understanding deposit type classification and aiding mineral exploration.

t-SNE visualizations showing the distribution of training samples and the 37 predicted samples from the Zawar deposit across the six deposit types using the RF (top row) and GB (bottom row) models.

---