Empirical characterization of random forest variable importance measures

By Kellie J. Archer, Ryan V. Kimes. 2008

1,070 citations recorded by the source corpus.

Subfields: Information Systems, Molecular Biology

How researchers see this paper

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

In Information Systems

In Molecular Biology

Citations in the corpus

All Models are Wrong, but Many are Useful: Learning a Variable’s Importance by Studying an Entire Class of Prediction Models Simultaneously (2018, methodology): “This RF VI measure has been the topic of empirical studies (Archer and Kimes, 2008; Calle and Urrea, 2010; Wang et al., 2016), and several variations of the measure have been proposed (Strobl et al.”
A new variable importance measure for random forests with missing data (2012, neutral): “A summary of several selection methods and related publications can be found in Archer and Kimes (2008).”
Zeileis Conditional Variable Importance for Random Forests (2015, methodology): “With respect to the identifiability of few influential predictors from a set of correlated and other noise variables (which was the task in [20] and [21]), we can see from the importance scores for X1,.”
A random forest guided tour (2015, background): “Indeed, Archer and Kimes (2008), Strobl et al. (2008), Nicodemus and Malley (2009), Auret and Aldrich (2011), and Toloşi and Lengauer (2011) stress the negative effect of correlated variables on MDA performance.”
Please Stop Permuting Features: An Explanation and Alternatives (2019, background): “Archer and Kimes (2008) explore a similar set-up and also note improved performance when true features – those actually related to the response – are uncorrelated with noise features.”
Random forest regression and spectral band selection for estimating sugarcane leaf nitrogen concentration using EO-1 Hyperion hyperspectral data (2013, background): “This error produces a measure of the importance of the variables by comparing how much OOB error of estimate increases when a variable is permuted, whilst all others are left unchanged (Archer and Kimes 2008; Verikas, Gelzinis and Bacauskiene, 2011).”
On Oblique Random Forests (2011, background): “Empirical studies suggest that both the feature rankings – the algorithmically motivated Gini importance, and the statistically defined permutation importance – correlate well in most classification tasks [1].”
A comparison of random forest and its Gini importance with standard chemometric methods for the feature selection and classification of spectral data (2009, background): “Inspecting the Gini feature importance we observed – similar to [32] – that some spectral regions were selected as a whole, suggesting that correlated variables were assigned similar importance.”
A balanced iterative random forest for gene selection from microarray data (2013, methodology): “Random forest has been used extensively in the biomedical domain [7,8] because it is well suited for microarray data.”
Ensemble Learning for Confidence Measures in Stereo Vision (2013, methodology): “Two different measures are used here for discussion [3]: The GINI importance measures the contribution of each variable to the decrease of the objective function, while the permutation importance calculates the decrease in accuracy on the out of bag samples after permuting the values of the feature.”
Computational Statistics and Data Analysis a Similarity Measure to Assess the Stability of Classification Trees (neutral): “A description and comparison of these measures can be found in Archer and Kimes (2008) and Strobl et al. (2007).”