Empirical comparison of tree ensemble variable importance measures

107 citations recorded by the source corpus.

How researchers see this paper

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

Consistent Individualized Feature Attribution for Tree Ensembles (2018, methodology): “Different choices about the method of feature value permutation lead to variations of this basic approach [1, 10, 14, 23, 26].”
A random forest guided tour (2015, background): “Similarly, the empirical study of Archer and Kimes (2008) points out that both MDA and MDI behave poorly when correlation increases—these results have been experimentally confirmed by Auret and Aldrich (2011) and Toloşi and Lengauer (2011).”
Correlation and variable importance in random forests (2013, background): “The eﬀect of the correlations on these measures has been studied in the last few years by Archer and Kimes [2], Strobl et al. [31], Nicodemus and Malley [27], Nicodemus et al. [28], Nicodemus [26], Auret and Aldrich [3] and Toloşi and Lengauer [34].”
Explainable AI for Trees: From Local Explanations to Global Understanding (2019, methodology): “The tuned gradient boosting model hyper-parameters were optimized over the tree depths of [1, 2, 4, 8, 10] and bagging sub-sampling over the rates [0.”