Lunch & Learn | Insurance risk classification with generalized gaussian process regression models

Don't miss our next Lunch & Learn on "Insurance risk classification with generalized gaussian process regression models". It's Free and Open to everyone.

Our Scientific Advisor, Donatien Hainaut, will present its Detra Note published in collaboration with Michel Denuit.

Why participate ?

In this Lunch & Learn, we present an actuarial regression model (GGPR), essentially non-parametric, which automatically manages the interactions present in the data. Generalized Gaussian Process Regressions (GGPR) are machine learning models with a Bayesian tone, suitable for predicting responses distributed according to one of the usual dispersed exponential laws in insurance.

In GGPR, individual random effects are distributed according to a multivariate normal law and connected using a covariance matrix constructed from a kernel function. This kernel imposes the regularity of predicted responses and ensures that they are essentially based on information borrowed from similar risk profiles.

Program

• Review of dispersed exponential distributions and combination with Gaussian Process Regression.
• Presentation of the calibration algorithm and main analytical results for usual dispersed exponential laws.
• Embedding of categorical variables and management of large databases through clustering.
• Case studies: auto (MTPL) and motorcycle (MCC) databases.

Actuary? This session is made for you!

💻 Free and Accredited training : 1 CPD | 6 PPC

🗓️ Join us online on Tuesday, May 20th from 12:30 to 1:30 PM.