Wiadomości

[Webinar OpenFloor] Transparent Models with Machine-Learning: Rethinking the Modeling and Validation Process – 30 March 2022

On behalf of Bruno Becha, Head of Actuarial Data Science at Akur8, the Board of the Polish Actuarial Association has the pleasure to invite its members and members of the other CEE actuarial associations to its complimentary „Open Floor” series webinar, „Transparent Models with Machine-Learning: Rethinking the Modeling and Validation Process.”

Abstract:

Actuarial modeling puts a strong emphasis on model transparency. Transparency and clarity is often obtained at the expense of the automation of the modeling process, and opposed to automated modeling or machine-learning.

In this presentation, we will discuss how transparency is a feature of the model, not the modeling process. With the right technologies, automated modeling can generate transparent models. We will demonstrate how credibility can be incorporated in standard GLM modeling by adding prior assumptions to the coefficients. Furthermore, we will explore how the choice of the right hypothesis can replace the need for heavy feature engineering on ordinal variables to develop sound and predictive models.

In particular, we will discover how a Generalized Linear Model with Credibility assumptions performs relative to classic criteria such as a likelihood or statistical testing, review the role of the modeler, and the model validation process, in an environment mixing automated model creation with an in-depth investigation of transparent models.

About the lecturer:

Bruno Becha is Head of Actuarial Data Science at Akur8. Bruno has been working as a Pricing Actuary and Data Scientist for the past 10 years, for insurance carriers and AI software companies. He has now been with Akur8 for 2 years as head of the Actuarial Data Science team, whose mission is to help Akur8’s clients accelerate and optimize their pricing process thanks to Transparent AI.

 

When: 30 March 2022 (Wednesday) at 17.00 CET, Zoom meeting. 

Duration: the webinar is expected to last 60-70 minutes, including the Q&A.

Technical remarks:

  • The lecture will be held in English.
  • The lecture will be held as webinar, rather than a meeting, therefore the participants won’t be able to turn on their cameras / microphone and share their screens with other attendees.
  • Questions can be asked via the Q&A box (only answered questions will be visible to all attendees).
  • The webinar is free of charge and no registration is required.