Feature Importance Scores rank inputs by what's?

Get ready for the GARP Risk and AI Exam with flashcards and multiple choice questions. Each question comes with hints and explanations. Prepare for success!

Feature importance scores rank inputs by how much they influence the model’s final predictions. They reflect the contribution each feature makes to the model’s ability to predict, such as how much error is reduced by using that feature or how much prediction gain comes from splits on that feature in tree-based models. Methods like permutation importance measure the drop in performance when a feature’s values are shuffled, while SHAP values assign a portion of each prediction to individual features. This focus on the feature’s impact on the output is why the correct choice is that feature importance reflects influence on the final model output.

Other options don’t fit because data collection difficulty and computational cost relate to practicality or resource use, not how much a feature affects predictions. Correlation with model accuracy can be tempting to think about, but a feature may be correlated with the target without actually adding useful information if its signal is redundant with other features.