Permutation importance: a corrected feature importance measure

By André Altmann, Laura Tolosi, Oliver Sander, Thomas Lengauer. 2010

2,570 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Molecular Biology, Evaluation & Benchmarks, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

In Evaluation & Benchmarks

In Safety, Robustness & Interpretability

Citations in the corpus

Explanations of model predictions with live and breakDown packages (2018, methodology): “Description of general framework of permutation based variable importance rankings may be found in Altmann et al. (2010). An interesting and widely adopted tool for estimation of marginal model response are Partial Dependency Plots (see Friedman (2001)), that presents the marginal relation between t”
All Models are Wrong, but Many are Useful: Learning a Variable’s Importance by Studying an Entire Class of Prediction Models Simultaneously (2018, methodology): “…are provided at http://jmlr.org/papers/v20/18-760.html. the decrease in prediction accuracy when a covariate is permuted (Breiman, 2001; Breiman et al., 2001; see also Strobl et al., 2008; Altmann et al., 2010; Zhu et al., 2015; Gregorutti et al., 2015; Datta et al., 2016; Gregorutti et al., 2017).”
County-Level Soybean Yield Prediction Using Deep CNN-LSTM Model (2019, methodology): “Permutation Feature Importance (PFI) is a commonly used method to evaluate the importance of the input variables [58].”
Explainable Artificial Intelligence Applications in Cyber Security: State-of-the-Art in Research (2022, methodology): “Local Interpretable Model-agnostic Explanations418 (LIME) [56] and Permutation Importance [57] are typical419 Post-hoc explanation methods working independently as an420 external interpretable model.421 2) MODEL-SPECIFIC OR MODEL-AGNOSTIC422 XAI methods can also be classified according to the classe”
What the Future Brings: Investigating the Impact of Lookahead for Incremental Neural TTS (2020, background): “the drop in R(2) that results from swapping columns in the dataset) [23].”
Metabolomics and Multi-Omics Integration: A Survey of Computational Methods and Resources (2020, background): “, Mean Decrease Accuracy [178] or Mean Decrease Gini [179,180]) when features are removed, or compare the influence of analytes on the model using the original values and shuffled values [181].”
All Models are Wrong but many are Useful: Variable Importance for Black-Box, Proprietary, or Misspecified Prediction Models, using Model Class Reliance (2018, methodology): “01 48 9v 3 the decrease in prediction accuracy when a covariate is permuted (Breiman, 2001; Breiman et al., 2001; see also Strobl et al., 2008; Altmann et al., 2010; Zhu et al., 2015; Gregorutti et al., 2015; Datta et al., 2016; Gregorutti et al., 2017).”
Physics-informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data (2016, background): “In the machine learning community, improving interpretability of random forest is an active research topic, e.g., several improvements of the importance measure have been proposed [55–57].”
Explainable AI: A Review of Machine Learning Interpretability Methods (2020, methodology): “As a result, an improved Random Forest (RF) model, called PIMP-RF, was proposed, which was trained on the most important features, as determined by the PIMP algorithm.”
Techniques for interpretable machine learning (2018, methodology): “A representative approach is Permutation Feature Importance [1].”
Towards a Science of Human-AI Decision Making: A Survey of Empirical Studies (2021, background): “followed by a rich line of research on the topic [4, 34, 56, 57, 131, 156].”