Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

University of Toronto

$27,000.00

Data-Driven Actuarial Modelling for General Insurance Ratemaking and Risk Management

FY 2023-24

Summary
My current research program develops probability and statistical models and methodology from actuarial science and insurance perspectives. The focus is on providing solutions to actuarial problems that have a real impact on insurance practice. This proposed research will focus on three closely related projects. 1. Mixture of Experts based models with applications in risk classification and ratemaking We explore applications of MoE type of models to insurance and in particular to risk classification and ratemaking. The  objectives are to develop a general framework for MoE models with multi-level, multi-factor random effects and apply it to posteriori risk classification and ratemaking. This model formulation will result in a flexible and highly nonlinear regression model and hence may provide better posteriori risk classification and ratemaking. The model developed above can be  useful for IBNR and RBNS predictions and the corresponding reserve calculations. 2. New approaches to credibility ratemaking with application to large insurance portfolios Experience/credibility ratemaking is referred to calculating insurance premiums for a policyholder based on the holder's claim history and attributes. Given a claim history, we can derive the predictive distribution of the future claim from the policyholder and the (credibility) premium. However, any Bayesian models that provide reasonable fits to real insurance data sets are mathematically intractable and the derivation of credibility premiums must rely on computationally expensive simulations. Our objective is to effectively calculate credibility premiums or other quantities of interest when we incorporate policyholders' claim history and attributes in a data-driven manner by using a sufficient summary statistic that may fully describe the claim history and policy attributes. A major advantage of this approach is its ability to scale up to the portfolio level via surrogate modelling, and thus provides useful and effective tools for risk managing insurance portfolios and the calculation of capital requirements. 3. Feature engineering, modelling and analysis of telematics data With the advancement in technology, vehicle movements can now be recorded by on-board diagnostics installed in the vehicle such as GPS locations, driving speed, acceleration and engine data, etc. and they are referred to as (vehicle) telematics data. The inclusion of telematics data will facilitate a more accurate ratemaking evaluation. The objective of this project is to focus on feature engineering using statistical models for accident prediction and to use machine learning methods for pattern recognition to see what the significant factors for accidents and the severity of the accidents are. The next phase is to consider functional regression models for insurance applications and commercial fleet risk management, and  user based insurance (UBI) for commercial fleets.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2023-04326
Name nom
Lin, Xiaodong
Department département
Statistical Sciences
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
27000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1202
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Physical sciences
Area of application fr
Sciences physiques
Research subject code
9999
Research subject group en
Not available
Research subject group fr
Not available
Research subject en
Non disponible
Research subject fr
Non disponible
Application title
Data-Driven Actuarial Modelling for General Insurance Ratemaking and Risk Management
Keyword
Actuarial Science; Insurance Ratemaking; Insurance Risk Managment; Property and Casualty Insurance; Reserving
Application summary
My current research program develops probability and statistical models and methodology from actuarial science and insurance perspectives. The focus is on providing solutions to actuarial problems that have a real impact on insurance practice. This proposed research will focus on three closely related projects. 1. Mixture of Experts based models with applications in risk classification and ratemaking We explore applications of MoE type of models to insurance and in particular to risk classification and ratemaking. The  objectives are to develop a general framework for MoE models with multi-level, multi-factor random effects and apply it to posteriori risk classification and ratemaking. This model formulation will result in a flexible and highly nonlinear regression model and hence may provide better posteriori risk classification and ratemaking. The model developed above can be  useful for IBNR and RBNS predictions and the corresponding reserve calculations. 2. New approaches to credibility ratemaking with application to large insurance portfolios Experience/credibility ratemaking is referred to calculating insurance premiums for a policyholder based on the holder's claim history and attributes. Given a claim history, we can derive the predictive distribution of the future claim from the policyholder and the (credibility) premium. However, any Bayesian models that provide reasonable fits to real insurance data sets are mathematically intractable and the derivation of credibility premiums must rely on computationally expensive simulations. Our objective is to effectively calculate credibility premiums or other quantities of interest when we incorporate policyholders' claim history and attributes in a data-driven manner by using a sufficient summary statistic that may fully describe the claim history and policy attributes. A major advantage of this approach is its ability to scale up to the portfolio level via surrogate modelling, and thus provides useful and effective tools for risk managing insurance portfolios and the calculation of capital requirements. 3. Feature engineering, modelling and analysis of telematics data With the advancement in technology, vehicle movements can now be recorded by on-board diagnostics installed in the vehicle such as GPS locations, driving speed, acceleration and engine data, etc. and they are referred to as (vehicle) telematics data. The inclusion of telematics data will facilitate a more accurate ratemaking evaluation. The objective of this project is to focus on feature engineering using statistical models for accident prediction and to use machine learning methods for pattern recognition to see what the significant factors for accidents and the severity of the accidents are. The next phase is to consider functional regression models for insurance applications and commercial fleet risk management, and  user based insurance (UBI) for commercial fleets.