Bias in Random Forest Variable Importance Measures¶
Strobl, C., Boulesteix, A., Zeileis, A., & Hothorn, T. (2007). Bias in Random Forest Variable Importance Measures: Illustrations, Sources and a Solution. BMC Bioinformatics.
Cited by¶
1 citation across 1 artifact.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Cross-Dimensional Leakage
- The corrective is experimental decoupling — permutation or interventional importance that breaks the confound — not a larger training set drawn through the same correlated feature distribution.
This sourceShows permutation/Gini variable-importance is biased toward features correlated with others, so a confounding correlated feature inflates the apparent importance of features that co-vary with it — and that breaking the dependence (conditional permutation) is required to resolve it.
- The corrective is experimental decoupling — permutation or interventional importance that breaks the confound — not a larger training set drawn through the same correlated feature distribution.
Verification¶
This reference passed the adversarial substantiation pipeline: it was checked to exist and to support the claim it is attached to. See how references were verified.
Registry ID ref:db2b41f0a6be · see in the full table