From global to local MDI variable importances for random forests and when they are Shapley values

Random forests have been widely used for their ability to provide so-called importance measures, which give insight at a global (per dataset) level on the relevance of input variables to predict a certain output. On the other hand, methods based on Shapley values have been introduced to refine the analysis of feature relevance in tree-based models to a local (per instance) level. In this context, we first show that the global Mean Decrease of Impurity (MDI) variable importance scores correspond to Shapley values under some conditions. Then, we derive a local MDI importance measure of variable relevance, which has a very natural connection with the global MDI measure and can be related to a new notion of local feature relevance. We further link local MDI importances with Shapley values and discuss them in the light of related measures from the literature. The measures are illustrated through experiments on several classification problems.

[1]  Antonio Sutera.  Importance measures derived from random forests: characterisation and extension , 2019 .

[2]  Wei-Yin Loh,et al.  Classification and regression trees , 2011, WIREs Data Mining Knowl. Discov..

[3]  Gaël Varoquaux,et al.  Scikit-learn: Machine Learning in Python , 2011, J. Mach. Learn. Res..

[4]  Alexey Ignatiev,et al.  On Explaining Decision Trees , 2020, ArXiv.

[5]  René Levínský,et al.  On proper Shapley values for monotone TU-games , 2015, Int. J. Game Theory.

[6]  Ron Kohavi,et al.  Wrappers for Feature Subset Selection , 1997, Artif. Intell..

[7]  Kellie J. Archer,et al.  Empirical characterization of random forest variable importance measures , 2008, Comput. Stat. Data Anal..

[8]  Erik Strumbelj,et al.  An Efficient Explanation of Individual Classifications using Game Theory , 2010, J. Mach. Learn. Res..

[9]  Gilles Louppe,et al.  Random Subspace with Trees for Feature Selection Under Memory Constraints , 2018, AISTATS.

[10]  H. Ishwaran.  Variable importance in binary regression trees and forests , 2007, 0711.2434.

[11]  Achim Zeileis,et al.  Bias in random forest variable importance measures: Illustrations, sources and a solution , 2007, BMC Bioinformatics.

[12]  H. Young.  Monotonic solutions of cooperative games , 1985 .

[13]  C. Aldrich,et al.  Empirical comparison of tree ensemble variable importance measures , 2011 .

[14]  Paola Zuccolotto,et al.  Variable Selection Using Random Forests , 2006 .

[15]  Manfred Besner.  Axiomatizations of the proportional Shapley value , 2017, Theory and Decision.

[16]  Gilles Louppe,et al.  Understanding variable importances in forests of randomized trees , 2013, NIPS.

[17]  Yu Wang,et al.  A Debiased MDI Feature Importance Measure for Random Forests , 2019, NeurIPS.

[18]  Scott Lundberg,et al.  A Unified Approach to Interpreting Model Predictions , 2017, NIPS.

[19]  Thomas M. Cover,et al.  Elements of Information Theory , 2005 .

[20]  Gilles Louppe,et al.  Understanding Random Forests: From Theory to Practice , 2014, 1407.7502.

[21]  Fernando V. Paulovich,et al.  Explainable Matrix - Visualization for Global and Local Interpretability of Random Forest Classification Ensembles , 2021, IEEE Transactions on Visualization and Computer Graphics.

[22]  Pierre Geurts,et al.  Extremely randomized trees , 2006, Machine Learning.

[23]  Erwan Scornet.  Trees, forests, and impurity-based variable importance , 2020 .

[24]  L. Shapley.  A Value for n-person Games , 1988 .

[25]  Scott Lundberg,et al.  Understanding Global Feature Contributions With Additive Importance Measures , 2020, NeurIPS.

[26]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[27]  Hugh Chen,et al.  From local explanations to global understanding with explainable AI for trees , 2020, Nature Machine Intelligence.