Improving road safety for heavy vehicle operations

Overview

Australia's National Heavy Vehicle Regulator (NHVR) used the Databricks Data + AI Platform to analyze more than 4.5 million monthly heavy-vehicle sightings and build machine learning models, including a fatigue engine and crash prediction model, to detect road safety risks in real time (subsecond reporting) across its national fleet of almost 1 million heavy vehicles. The fatigue engine delivered a 4% increase in yielding high-risk fatigue infringements, and the crash model identified vehicles/operators with a 1-in-38 chance of a fatal or serious incident on any given day.

The challenge

NHVR had developed a data acquisition strategy to better understand the industry it regulates, and, as a result, started to collect large amounts of data sets that its warehouse could not ingest quickly enough. Limited by the capabilities of its traditional data warehousing solution, the NHVR needed a platform that supported real-time processing cost-effectively, to enable predictive analysis and faster time to insight, in order to disrupt in-flight safety threats.

The solution

The NHVR turned to Databricks Data + AI Platform, which allows its data science team to run machine learning experiments quickly and easily. The NHVR built a fatigue engine to help identify and target drivers and operators most at risk of causing fatigue-related accidents, and created a more reliable crash prediction model powered by machine learning, leveraging multiple data sets including vehicle data, defect data and crash data.

Reported business value

This has resulted in a 4% increase in yielding high-risk fatigue infringements. The crash prediction model was able to identify a cohort of vehicles and operators that had a 1-in-38 chance of being involved in a fatal or serious incident on any given day, an actionable prediction that lets NHVR deploy safety and compliance officers to target these specific vehicles.