Our client, a chemicals company, owns and operates a large trucking fleet. They desire smarter and more automated ways to keep their drivers safer, through predictive analytics. Armed with new driver and vehicle data, we hosted an algorithm challenge that delivered impressive results.
Existing methodologies for linking historical data to fleet safety have been helpful in identifying high-risk routes and drivers. However, efficient algorithms, driven by multiple factors like weather patterns, routes, vehicle specificity, driving patterns, specific payload transport risk, and other factors, had not been robustly developed with the purpose of identifying indicators that could be used to more precisely measure risk and increase safety.
With the proliferation of on-board computers (OBC) and sensors that track actions like braking patterns and vehicle behavior, our client wanted to create an algorithmic solution that could ultimately help keep their fleet of drivers, and those they share the road with, safer.
A new on-board computer (OBC-based) trip risk prediction metric was created and provided a valid way to test algorithmic solutions against, to measure precision of the submitted solutions.
Over 1000 algorithmic solutions from over 100 registrants to this challenge were procured during the 2-week long challenge period.
The winning solutions delivered a precision improvement of > 40% as compared to the baseline algorithm which had a precision scoring of < 5%.