Interpretation & discovery
Feature explanations and SHAP
Understand Ounias feature-attribution outputs and their limitations.
Overview
Feature explanations help identify which variables contribute most strongly to the analytical result.
Ounias uses SHAP-based explanations and keeps the explanation artifacts attached to the run that produced them.
Available outputs
- Feature importance
- Run-linked SHAP artifacts
- Feature-level contribution information
- Comparison with other analytical outputs
- Exportable explanation tables
Interpretation and limitations
- Feature importance describes model behavior rather than causal influence.
- Correlated variables can complicate attribution.
- Explanation quality depends on the model, data representation, and analytical configuration.