Colloquium: Quantifying Driving Risk via Telematics Data: a Wavelet Transform Approach

TORONTO METROPOLITAN UNIVERSITY DEPARTMENT OF MATHEMATICS COLLOQUIUM

Dr. Sheldon Lin

Department of Statistical Sciences University of Toronto

Quantifying Driving Risk via Telematics Data: a Wavelet Transform Approach

Abstract: Vehicle telematics provides granular data for dynamic driving risk assessment, but most existing methods to assess driving risk rely on aggregated metrics/summary statistics. As a result, they do not fully exploit the rich time-series structure of trip data. In this project, we propose the use of wavelet transform to capture key driving patterns of trips which in turn are used to evaluate driver and trip-level behaviors. By incorporating both frequency and severity of acceleration events, this approach enables to distinguish aggressive and inattentive driving from normal driving behavior, as demonstrated in a controlled study. Furthermore, the approach enhances interpretability, reduces misclassification, and provides a flexible foundation for telematics-based risk assessment. This is joint work with Andrei Badescu and Jongtaek Lee.