Discover how we compare the road collision prediction performance level of two analyses: the first one is conducted based on historical road collisions and the other is based on harsh braking data only. We quantify the benefits in terms of accuracy and speed when it comes to predicting future collisions on a large-scale road network using real data from fleets of connected vehicles. We demonstrate that harsh braking clusters enable us to detect precise spots that are dangerous for road users.

The study demonstrates that clusters of harsh braking events:

• Bring comparatively more information than past collisions (22% conversion rate vs 6% with 5 years of KSI data).

• Enable an improvement of 22% in prediction rate compared to an equivalent number of historical crash locations.

• Allow much quicker detection of future collisions.

• Enabled us to detect collisions that could not be predicted even using 5 years of historical road crash data.

This research demonstrates that road safety engineers and managers now have access to superior tools to enhance road infrastructure and make more informed, risk-based decisions.