Mean decrease accuracy for random forests: inconsistency, and a practical solution via the Sobol-MDA

DOI10.1093/BIOMET/ASAC017MaRDI QIDQ5872839zbMATH OpenOpenAlexWikidataFDO

Authors Clément Bénard, Sébastien Da Veiga, Erwan Scornet

Publication date 4 January 2023

Published in Biometrika (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1093/biomet/asac017

zbMATH Keywords

variable importancevariable selectionsensitivity analysisrandom forestMDASobol index

Mathematics Subject Classification ID

Statistics (62-XX)

Recommendations

• A unified framework for random forest prediction error estimation
• Variance reduction in purely random forests
• Consistency of random forests and other averaging classifiers
• Towards convergence rate analysis of random forests for classification
• Consistent estimation of residual variance with random forest out-of-bag errors
• Models under which random forests perform badly; consequences for applications
• Random Forest Adjustment for Approximate Bayesian Computation

Cited in

(3)

• Quantile-constrained Wasserstein projections for robust interpretability of numerical and machine learning models
• Feature importance: a closer look at Shapley values and LOCO
• Total effects with constrained features

This page was built for publication: Mean decrease accuracy for random forests: inconsistency, and a practical solution via the Sobol-MDA

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5872839)