A service provided by the WU Library and the WU IT-Services

Conditional Variable Importance for Random Forests

Strobl, Carolin and Boulesteix, Anne-Laure and Kneib, Thomas and Augustin, Thomas and Zeileis, Achim (2008) Conditional Variable Importance for Random Forests. BMC Bioinformatics, 9 (307). pp. 1-11. ISSN 1471-2105

[img]
Preview
PDF
Download (440Kb) | Preview

Abstract

Background Random forests are becoming increasingly popular in many scientific fields because they can cope with "small n large p" problems, complex interactions and even highly correlated predictor variables. Their variable importance measures have recently been suggested as screening tools for, e.g., gene expression studies. However, these variable importance measures show a bias towards correlated predictor variables. Results We identify two mechanisms responsible for this finding: (i) A preference for the selection of correlated predictors in the tree building process and (ii) an additional advantage for correlated predictor variables induced by the unconditional permutation scheme that is employed in the computation of the variable importance measure. Based on these considerations we develop a new, conditional permutation scheme for the computation of the variable importance measure. Conclusion The resulting conditional variable importance reflects the true impact of each predictor variable more reliably than the original marginal approach. (authors' abstract)

Item Type: Article
Additional Information: CS defined the research question, suggested the conditional variable importance, set up and performed the simulation experiments and drafted the manuscript. A-LB analyzed the peptide-binding data. TK, TA and AZ contributed to the theoretical understanding and presentation of the problem. All authors contributed to and approved the final version of the manuscript.
Divisions: Departments > Finance, Accounting and Statistics > Statistics and Mathematics
Version of the Document: Published
Variance from Published Version: None
Depositing User: Elena Simukovic
Date Deposited: 14 Apr 2016 12:05
Last Modified: 19 Apr 2016 13:27
Related URLs:
FIDES Link: https://bach.wu.ac.at/d/research/results/44072/
URI: http://epub.wu.ac.at/id/eprint/4987

Actions

View Item