Response versus gradient boosting trees, GLMs and neural networks under Tweedie loss and log-link

Donatien Hainaut, Julien Trufin and Michel Denuit
Additional contact information
Donatien Hainaut: Université catholique de Louvain, LIDAM/ISBA, Belgium
Julien Trufin: Université Libre de Bruxelles
Michel Denuit: Université catholique de Louvain, LIDAM/ISBA, Belgium

No 2021012, LIDAM Discussion Papers ISBA from Université catholique de Louvain, Institute of Statistics, Biostatistics and Actuarial Sciences (ISBA)

Abstract: Thanks to its outstanding performances, boosting has rapidly gained wide acceptance among actuaries. To speed calculations, boosting is often applied to gradients of the loss function, not to responses (hence the name gradient boosting). When the model is trained by minimizing Poisson deviance, this amounts to apply the least-squares principle to raw residuals. This exposes gradient boosting to the same problems that lead to replace least-squares with Poisson GLM to analyze low counts (typically, the number of reported claims at policy level in personal lines). This paper shows that boosting can be conducted directly on the response under Tweedie loss function and log-link, by adapting the weights at each step. Numerical illustrations demonstrate improved performances compared to gradient boosting when trees, GLMs and neural networks are used as weak learners.

Keywords: Risk classification; Boosting; Gradient Boosting; Regression Trees; GLM; Neural Networks (search for similar items in EconPapers)
Pages: 26
Date: 2021-01-01
New Economics Papers: this item is included in nep-big and nep-cmp
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://dial.uclouvain.be/pr/boreal/fr/object/bore ... tastream/PDF_01/view (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:aiz:louvad:2021012

Access Statistics for this paper

More papers in LIDAM Discussion Papers ISBA from Université catholique de Louvain, Institute of Statistics, Biostatistics and Actuarial Sciences (ISBA) Voie du Roman Pays 20, 1348 Louvain-la-Neuve (Belgium). Contact information at EDIRC.
Bibliographic data for series maintained by Nadja Peiffer ().