XAI in non-life insurance pricing + workshop

The January meeting will be divided into two parts, workshop (we will share the details soon) and Michael Mayer's presentation entitled "XAI in non-life insurance pricing".

Talk abstract:

We will cover two topics. Firstly, I will give you an overview about statistical modelling in non-life insurance pricing and its specialties. How can modern machine learning challenge the good old GLM world? What is the role of ML interpretability and explainability in this challenge? I will illustrate the ideas on a large real world motor claims data set.

In the second part I’d like to share some thoughts on model-agnostic measures of interaction strength. How can they be used to efficiently measure overall interaction strength per input variable? How can shortcomings in current partial dependence related measures be mended?

Bio:

Michael Mayer is an enthusiastic statistician and R programmer working as non-life pricing actuary for “Die Mobiliar”, one of the most traditional and successful insurance companies in Switzerland. Since finishing his PhD at the University of Bern in 2008, Michael has worked as biostatistician, predictive modeler, statistics consultant as well as University lecturer. He is the author of three CRAN packages.