Features‎ > ‎

Model selection

Transmin recommends the following golden rules for geometallurgical model selection:

  1. Correlation.  The model should correlate to the data
  2. Explainable.  Geologists and metallurgists should have an explanation of why the model works
  3. Parsimony.  The model should be as simple as possible
  4. Contrast.  The model should show a contrast between values
  5. Projectable.  The model should apply to all of the blocks in the mine model
  6. Useful.  The model should use features that can be confidently populated to the block model
  7. Continuity.  The model should show spatial continuity over volumes that are useful for mine planning
  8. Significant.  The model should have statistical significance
Cancha automates the process:
  1. Geometallurgical model generation
  2. Statistical validation
  3. Model ranking 
While models are ranked and recommended, model validation and selection should be done by the user.  

With little preparatory work by the user, Cancha rapidly generates and ranks models from thousands of data features and sample characteristics, such as:
  • Head assays
  • Automated modal mineralogy
  • Lithology and alteration rock types
  • Geological logs
  • Spatial location
Model reports include statistics and graphs on correlations, features, and residuals to aid the user in model evaluation and selection.