This post explains the tidymodels blackbox with various explainability methods such as permutation importance, partial dependence plots, Friedman's H statistics, and SHAP. It uses the diabetes prediction dataset to model diabetes as a function of demographic features. The post also introduces the own packages 'hstats', 'kernelshap', and 'shapviz' for explanation methods.
•5m read time• From r-bloggers.com
Sort: