.png){kind=icon}
.png){kind=icon}
Return
A machine learning approach in type-discrimination and exploration of gold deposits using pyrite trace element chemistry
.png){kind=icon}
PRE
AI
A
R
D
S
K
A
DOI：10.1016/j.gexplo.2026.108011.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The trace element composition of pyrite can provide valuable insights into the genesis of ore deposits. However, discriminating between deposit types still remains challenging due to overlapping trace element signatures. This study uses advanced machine learning techniques to discriminate pyrite from six different types of gold deposits viz., Carlin, high sulfidation epithermal (HSE), iron oxide copper-gold (IOCG), low sulfidation epithermal (LSE), orogenic, and porphyry types, using its trace element composition. A large dataset, consisting of 8598 pyrite analyses from 112 deposits was compiled. Three state-of-the-art machine learning (ML) algorithms-XGBoost, LightGBM, and CatBoost - were applied to classify deposit types based on eleven trace elements. All the three models provide high accuracy (>94%) in deposit types discrimination. SHAP (SHapley Additive exPlanations) analysis was performed to identify key discriminating elements for each deposit type. The most influential elements controlling the discrimination are Cu, Co, and As. We also built ML models to discriminate barren sedimentary pyrite from mineralized pyrite, achieving >97% classification accuracy across all three models. Gold, Sb, and Se were identified as the key discriminators in this case. Pyrite associated with gold mineralization contains higher concentration of Au, and Sb, reflecting hydrothermal enrichment, whereas barren sedimentary pyrite typically exhibits higher Se concentrations. We developed a user-friendly web application that allows realtime classification of gold deposits from pyrite trace elements data. The efficacy of the models is demonstrated by case studies on the Pianyanzi orogenic gold deposit and the Shuanglong IOCG deposit. Our ML classifier models can be used as a robust tool for gold exploration and deposit types characterization using pyrite geochemistry.
Keywords:
Pyrite
Trace elements
Gold deposits
Machine learning
Deposit classification
SHAP analysis
Mineral exploration
Journal
IF:
3.3
Papers:
3.8K
Citations:
9.5K
Organization
Cited Papers