Interpreting uninterpretable predictors: kernel methods, Shtarkov solutions, and random forests. Issue 1 (2nd January 2022)

• Record Type: Journal Article
• Authors: Le, T. M.; Clarke, Bertrand
• Is Part Of: Statistical theory and related fields. Volume 6:Issue 1(2022)
• Page Start: 10
• Page End: 28
• Publication Date: 2022-01-02
• Subjects: Bayes -- boosting -- kernel methods -- random forest -- Shtarkov predictor -- stacking; Statistics -- Periodicals; Statistics; Periodicals; Electronic journals; 001.422
• DOI: 10.1080/24754269.2021.1974157
• Languages: English
• ISSNs: 2475-4269

Abstract

Abstract : Many of the best predictors for complex problems are typically regarded as hard to interpret physically. These include kernel methods, Shtarkov solutions, and random forests. We show that, despite the inability to interpret these three predictors to infinite precision, they can be asymptotically approximated and admit conceptual interpretations in terms of their mathematical/statistical properties. The resulting expressions can be in terms of polynomials, basis elements, or other functions that an analyst may regard as interpretable.