Feature importances Machine Learning
noun phrase
Definition: Quantitative scores that indicate how much each feature contributes to a model’s predictions or predictive performance, commonly used in tree-based models and other interpretability workflows [scikit-learn documentation].
Example in context: “TabNet translates the local and global interpretability as feature importances.” [Borsos et al. 2023]
Synonym: variable importances
Related terms: feature importance scores, permutation importance, SHAP values (related interpretability method)