Chapter 5: Feature Importance

Feature importance (FI) methods rank features by how much they contribute to the predictive performance or prediction variance of the model. With a loss-based focus this section deals with the FI methods PFI, CFI, SAGE and LOCO.