This research explores the spatial and temporal correlations between unsafe driving behaviours ,characterised by harsh braking, rapid acceleration, and other telematics-based events, and traffic crashes at 439 urban intersections in central Athens. Utilizing a comprehensive dataset comprising telematics data from 2019, gathered via smartphone applications, and multi-year crash records obtained from police reports, this study integrates these sources to classify intersections by crash risk and delineate high-risk “danger zones.” Employing advanced machine learning techniques, including Random Forest, XGBoost, Gradient Boosting Machines, and Logistic Regression, the analysis rigorously evaluates the predictive capacity of telematics-derived metrics in forecasting crash risks. The study leverages cutting-edge geospatial analytics to uncover critical patterns that link dynamic driving behaviours to elevated crash probabilities. These findings not only advance the methodological framework for crash risk prediction but also provide actionable insights for urban traffic safety management. By enabling proactive identification of high-risk intersections, this research contributes to the design of targeted, data-driven interventions aimed at mitigating crash risks and enhancing road safety in complex urban environments.
| ID | pc607 |
| Presentation | |
| Full Text | |
| Tags |
