intelligent mineral prospectivity mapping based on multi-view graph learning

Key Points

• The aim is to develop a robust method for mineral prospectivity mapping using diverse geological data.
• Constructed multi-view graph structures from spatial, geochemical, and geological data.
• Incorporated self-supervised learning constraints within a residual GraphSAGE framework.
• Employed a semi-supervised learning approach utilizing limited labeled samples.
• The proposed method outperformed comparative models in AUPRC and AUROC metrics.
• Predicted results showed strong spatial consistency with major mining structures and geochemistry.
• Indicates a high application potential for evaluating mineral prospectivity in weakly annotated regions.