Understanding overfitting in random forest for probability estimation: a visualization and simulation study
Random forests learn local probability peaks that often yield near perfect training AUCs without strongly affecting AUCs on test data. When the aim is probability estimation, the simulation results go against the common recommendation to use fully grown trees in random forest models.
Authors
- D. Timmerman (ORCID: https://orcid.org/0000-0002-3707-6645)
- Anne‐Laure Boulesteix (ORCID: https://orcid.org/0000-0002-2729-0947)
- Lasai Barreñada (ORCID: https://orcid.org/0000-0001-8020-0210)
- Paula Dhiman (ORCID: https://orcid.org/0000-0002-0989-0623)
- Ben Van Calster (ORCID: https://orcid.org/0000-0003-1613-7450)
Institutions
- Zimmer Biomet (Netherlands) (NL)
- Leiden University Medical Center (NL)
- Nuffield Orthopaedic Centre (GB)
- University of Oxford (GB)
- VIB-KU Leuven Center for Microbiology (BE)
- Ludwig-Maximilians-Universität München (DE)
- KU Leuven (BE)
Publication Details
- Journal
- Diagnostic and Prognostic Research
- Published
- 2024-09-27
- DOI
- https://doi.org/10.1186/s41512-024-00177-1
- Citations
- 94
- Primary Topic
- Gaussian Processes and Bayesian Inference
- Type
- article
- Field-Weighted Citation Impact
- 32.65
Funders
- Fonds Wetenschappelijk Onderzoek
- Vlaamse regering