Integration of traditional and telematics data via mass imputation

While telematics data with measurements of specific driving behaviors such as speeding, breaking, and turning, have shown significant predictive power in modeling accident risk, its usability has been challenged due to the low proportion of auto insurance policyholders that agree to provide their telematics data to the insurance companies. In this regard, we propose a novel method to create the unobserved telematics features out of the observed traditional features and insurance claim experience based on the so-called mass-imputation method for handling both telematics data scarcity and selection bias. It is also expected that by blending methodological expertise from the home institution with the host’s access to Canadian insurance applications, this partnership will set the stage for joint student supervision, co-authored publications, and mutual visiting scholar opportunities.

Faculty Supervisor:

Himchan Jeong

Student:

Partner:

Yonsei University

Discipline:

Mathematics

Sector:

Education

University:

Simon Fraser University

Program: