TY - JOUR
T1 - Model-agnostic variable importance for predictive uncertainty
T2 - an entropy-based approach
AU - Wood, Danny
AU - Papamarkou, Theodore
AU - Benatan, Matt
AU - Allmendinger, Richard
N1 - Publisher Copyright:
© The Author(s) 2024.
PY - 2024/8/29
Y1 - 2024/8/29
N2 - In order to trust the predictions of a machine learning algorithm, it is necessary to understand the factors that contribute to those predictions. In the case of probabilistic and uncertainty-aware models, it is necessary to understand not only the reasons for the predictions themselves, but also the model's level of confidence in those predictions. In this paper, we show how existing methods in explainability can be extended to uncertainty-aware models and how such extensions can be used to understand the sources of uncertainty in a model's predictive distribution. In particular, by adapting permutation feature importance, partial dependence plots, and individual conditional expectation plots, we demonstrate that novel insights into model behaviour may be obtained and that these methods can be used to measure the impact of features on both the entropy of the predictive distribution and the log-likelihood of the ground truth labels under that distribution. With experiments using both synthetic and real-world data, we demonstrate the utility of these approaches in understanding both the sources of uncertainty and their impact on model performance.
AB - In order to trust the predictions of a machine learning algorithm, it is necessary to understand the factors that contribute to those predictions. In the case of probabilistic and uncertainty-aware models, it is necessary to understand not only the reasons for the predictions themselves, but also the model's level of confidence in those predictions. In this paper, we show how existing methods in explainability can be extended to uncertainty-aware models and how such extensions can be used to understand the sources of uncertainty in a model's predictive distribution. In particular, by adapting permutation feature importance, partial dependence plots, and individual conditional expectation plots, we demonstrate that novel insights into model behaviour may be obtained and that these methods can be used to measure the impact of features on both the entropy of the predictive distribution and the log-likelihood of the ground truth labels under that distribution. With experiments using both synthetic and real-world data, we demonstrate the utility of these approaches in understanding both the sources of uncertainty and their impact on model performance.
KW - Entropy
KW - feature importance
KW - predictive uncertainty
KW - variable importance
U2 - 10.1007/s10618-024-01070-7
DO - 10.1007/s10618-024-01070-7
M3 - Article
SN - 1384-5810
JO - Data Mining and Knowledge Discovery
JF - Data Mining and Knowledge Discovery
ER -