Visiting Speaker: Dr Christopher Blier-Wong on Representation Learning for Insurance Pricing

Friday, Feb. 20, 1-2:30 p.m.

Room BN-3007, Business Administration building, St. John's campus, Memorial University

Dr Christopher Blier-Wong, Assistant Professor, Department of Statistical Sciences, University of Toronto (https://www.statistics.utoronto.ca/people/directories/all-faculty/christopher-blier-wong)

Three applications of representation learning for insurance pricing

Dr Blier-Wong’s (research interests lie in the intersection of actuarial science, machine learning, and statistics, with a particular focus on dependence modelling and e-variables.

Accurate insurance pricing depends on how well risk information is represented in the data. This talk shows how to build improved representations that represent context, non-linearities, and interactions, without changing the goal of pricing: better prediction and clearer modelling choices. In this talk, Dr Blier-Wong will will discuss three recent projects that center on this core idea:

Creation of location summaries from satellite imagery and OpenStreetMap features so that each policy is linked to a compact description of its surrounding environment.
Connection of flexible tree-based prediction to interpretable regression by showing that a generalized linear model can reproduce random-forest-fitted predictions exactly when given variables derived from the trees’ internal splits.
Use of large language models to convert policyholder text into numerical representations that represent semantic meaning and context, improving predictions when numerical variables are expressed in textual form.

This talk is open to all students, faculty, staff and the public.

See https://www.mun.ca/economics/research/speaker-series/

Presented by Department of Economics

Event Listing 2026-02-20 13:00:00 2026-02-20 14:30:00 America/St_Johns Visiting Speaker: Dr Christopher Blier-Wong on Representation Learning for Insurance Pricing Dr Christopher Blier-Wong, Assistant Professor, Department of Statistical Sciences, University of Toronto (https://www.statistics.utoronto.ca/people/directories/all-faculty/christopher-blier-wong) Three applications of representation learning for insurance pricing Dr Blier-Wong’s (research interests lie in the intersection of actuarial science, machine learning, and statistics, with a particular focus on dependence modelling and e-variables. Accurate insurance pricing depends on how well risk information is represented in the data. This talk shows how to build improved representations that represent context, non-linearities, and interactions, without changing the goal of pricing: better prediction and clearer modelling choices. In this talk, Dr Blier-Wong will will discuss three recent projects that center on this core idea: Creation of location summaries from satellite imagery and OpenStreetMap features so that each policy is linked to a compact description of its surrounding environment. Connection of flexible tree-based prediction to interpretable regression by showing that a generalized linear model can reproduce random-forest-fitted predictions exactly when given variables derived from the trees’ internal splits. Use of large language models to convert policyholder text into numerical representations that represent semantic meaning and context, improving predictions when numerical variables are expressed in textual form. This talk is open to all students, faculty, staff and the public. See https://www.mun.ca/economics/research/speaker-series/ Room BN-3007, Business Administration building, St. John's campus, Memorial University Department of Economics