| English | 中文 |
Use the interpretation method to decide the population, comparison and uncertainty the explanation represents. Contributions describe a fitted prediction model; they do not identify causal effects.
The implemented R module contribution/explain.R runs validation permutation importance, training-background PDP/ICE and optional empirical-background Shapley estimates. It shares the fitted bundle and common preprocessing; it is not a separate CLI or another fitting pipeline. Configure explanation in the existing R task. R settings retain exact sample, feature, grid and computation bounds.
Run through the R operations, then inspect requested explanation roles via artifacts.json, background/sample choices, units, categorical levels, dependence and uncertainty. Optional output absence must match the requested settings. Python prediction availability does not imply Python explanation support; check the Python path.