Detra Note 2023-3 | Boosted Poisson regression trees: A guide to the BT package in R18/04/2023
Detra Note 2023-4 | Cyber incident reports: extrapolating severity using neural networks29/05/2023
Happy Learning Hours | Insurance Pricing : a walk through recent results on autocalibration
We are happy to invite you to participate in our second Happy Learning Hours on ”Insurance Pricing”.
What are the Happy Learning Hours ? It is a FREE after-work learning session given by our Team members and followed by a drink. You will be accredited of 2 CPD points.
This session will take place on Tuesday June 27, 2023, from 17:30 to 19:30 at The Lincoln Auditorium (Brussels Manhattan Center - 3d floor).
It will be given in French (with support in English) by Michel Denuit and Julien Trufin (Scientific Advisors at Detralytics and renowned Professor of Actuarial Science).
ABOUT THE SESSION
By exploiting massive amounts of data, machine learning techniques provide actuaries with predictors exhibiting high correlation with claim frequencies and severities. However, these predictors generally fail to achieve financial equilibrium and thus do not qualify as pure premiums. Autocalibration effectively addresses this issue since it ensures that every group of policyholders paying the same premium is on average self-financing.
During this walk, we propose to look at four recent results concerning autocalibration.
1. First, we establish a new characterization of autocalibration, based on the graphs of Lorenz and Concentration curves. We exploit this result to propose an effective testing procedure for auto-calibration.
2. Then, we study the effect of balance correction on resulting pure premiums. We show that this method is beneficial in terms of out-of-sample, or predictive Tweedie deviance, Bregman divergence as well as concentration curves.
3. After that, we explore performances criteria in the class of autocalibrated predictors. We show that the Gini index and the elementary Pearson’s correlation are consistent scoring rules in the class of autocalibrated predictors.
4. Finally, we explore Tweedie dominance to compare competing models. A necessary and sufficient condition is established under autocalibration.
Address : Brussels Manhattan Center (3d floor) : Av. du Boulevard 21 - 1210 Brussels
Access : Please check in at the Manhattan Center front desk to obtain your visitor badge and access the 3d floor.
Parking : Manhattan Parking – Indigo : Rue des Croisades, 19 – 1210 Saint-Josse-Ten-Noode (with direct access to the Manhattan Center)
Metro : Rogier (with direct access to the Manhattan Center)