Tarification des risques en assurance non-vie, une approche par modèle d’apprentissage statistique

Antoine Pagliat () and Martial Phélippé-Guinvarc'H ()
Additional contact information
Antoine Pagliat: Groupama - Agricultural Insurance Service of GROUPAMA - Agricultural Insurance Service of GROUPAMA
Martial Phélippé-Guinvarc'H: Groupama - Agricultural Insurance Service of GROUPAMA - Agricultural Insurance Service of GROUPAMA

Post-Print from HAL

Abstract: Non-life actuarial researches mainly focus on improving Generalized Linear Models. Nevertheless, this type of model sets constraints on the risk structure and on the interactions between explanatory variables. Then, a bias between the real risk and the predicted risk by the model is often observed on a part of data. Nonparametric tools such as machine learning algorithms are more efficient to explain the singularity of the policyholder. Among these models, regression trees offer the benefit of both reducing the bias and improving the readability of the results of the pricing estimation. Our study introduces a modification of the Classification And Regression Tree (CART) algorithm to take into account the specificities of insurance data-sets. It compares the results produced by this algorithm to these obtained using Generalized Linear Models. These two approaches are then applied to the pricing of a vehicle insurance portfolio.

Date: 2011-12-01
References: Add references at CitEc
Citations:

Published in Bulletin français d'actuariat, 2011, 11, pp.49 - 81

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

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:hal:journl:hal-02151827

Access Statistics for this paper

More papers in Post-Print from HAL
Bibliographic data for series maintained by CCSD ().