group-SHAP uncertainty
Published in SOIL, 2024
Recommended citation: Rohmer, J., Belbeze, S., and Guyonnet, D.: Insights into the prediction uncertainty of machine-learning-based digital soil mapping through a local attribution approach, SOIL, 10, 679–697, https://doi.org/10.5194/soil-10-679-2024, 2024. https://soil.copernicus.org/articles/10/679/2024/
The present work proposes a simple but effective approach based on a popular method base don the SHAP approach. The novelty is its application here to better understand which covariates affect the most the uncertainty of the prediction, i.e. to explain why the ML model is reliable, given the set of covariate values chosen.
Recommended citation: Rohmer, J., Belbeze, S., and Guyonnet, D.: Insights into the prediction uncertainty of machine-learning-based digital soil mapping through a local attribution approach, SOIL, 10, 679–697, https://doi.org/10.5194/soil-10-679-2024, 2024.