Spatio-Temporal Models for the Analysis of GPS Traces: Application to Road Safety
The goal of this research is to leverage the telematics data collected in the context of the usage-based insurance program at Intact for road safety improvement. Specifically, we aim to tackle issues on the identification of risky driver behaviors through the characterization of unsafe events and to identify sites on the road network with high probability of collision. The objectives of this research project are defined around three research themes: 1) Automated detection and characterization of unsafe events at the driver level using unsupervised, semi- supervised and supervised machine learning models; 2) Computation of contextual variables at the network level (traffic and congestion measures); and 3) Network screening and validation of the measures defined in themes 1) and 2). This project will revolve around the third theme which will involve the development of spatio-temporal point process models for network screening using historical crash database. This project is part of a larger research project carried at HEC Montreal on road safety in collaboration with several researchers from McGill, Polytechnique and HEC Montreal. By developing a better framework and better tools for GPS and motion sensor data analysis, Intact Insurance will gain a competitive edge over its competitors, both nationally and internationally.