Curbing traffic accidents and saving lives with machine learning

Artificial intelligence and cloud computing make roads safer in Western Australia

Western Australia can be a challenging place to drive for locals and visitors alike. If the state separated from the rest of Australia, it would rank as the 10^th largest country in the world. The sheer scale and vast distances involved in traveling throughout Western Australia make road safety a particularly challenging issue. While efforts to improve road safety have curtailed the number of casualties and serious injuries over the past decade, the state is still behind the Australian national average. 

Consider the beautiful and increasingly popular Indian Ocean Drive. This picturesque highway, which hugs Australia’s west coast from Perth to Arrowsmith, was recently the site of 50 serious road accidents in just 18 months – signaling an alarming spike in road trauma that’s created a swirl of media attention and left officials scrambling for a proactive solution.

Enter the Road Safety Commission of Western Australia. As a business unit of the Western Australia Police Force, the Commission is charged with tackling road trauma, which ranks as one of the biggest and most sustained causes of death and injury in the state. It does so by incorporating input from the private sector, government agencies and the community, and using it to inform policy recommendations and mass media campaigns.

The Commission has guided the implementation of Western Australia’s road safety strategy, “Towards Zero 2008-2020,” which aims to improve road safety and achieve a 40% reduction in road fatalities and serious injuries. Previously, the Commission relied heavily on human intelligence to translate insight into action. But in the wake of the troubling trends highlighted by Indian Ocean Drive, it’s increasingly using artificial intelligence (AI) and advanced analytics to bolster safety efforts. 

A reactive approach wasn’t working

David Slack-Smith is Manager of Data and Intelligence at the Road Safety Commission. For this dedicated public servant – who transitioned from the Australian Army to the police force – SAS has been instrumental in changing the manner and speed in which his team works.

His team’s role is to provide the statistical evidence that ensures all funding the Commission receives is well spent. An example of this involves the placement of red-light speed cameras, a proven and effective way to stop crashes at intersections. To achieve the best possible placement, his small yet agile team of analysts previously relied on manually ranking road intersections using killed or seriously injured (KSI) statistics.

But they found several challenges with this approach. First, his team discovered that a weighted average doesn’t adequately assess risk. The process was also labor-intensive, taking up to 100 hours to analyze historical crash data and perform backcasting. Finally, the team’s use of Microsoft Excel made version control and governance difficult.

“When the goal is to eliminate road trauma, it can’t be achieved by being reactive, if you are looking in the rear-view mirror,” Slack-Smith says. Clearly, a change was needed to replace the cumbersome and reactive process with a proactive solution to curbing road accidents.

How cloud AI accelerates speed, accuracy and efficiency

The Commission invested in SAS Visual Data Mining and Machine Learning powered by SAS Viya on SAS Cloud. The solution enabled the rapid prototyping of a machine learning model to produce a probability of KSI crashes, enabling the team to be forward-looking by offering better insight for devising safety measures and offering several key advantages.

Fast deployment

Using the same data inputs, as well as new variables including road features and traffic volumes, Slack-Smith and his team deployed the highly innovative model in just three weeks. Now the complete analytics lifecycle for a project – from data engineering to data visualization for stakeholders – has dropped from 100 hours to just 20 hours.

Better accuracy

Not only is the SAS machine learning model five times faster than the previous method, it’s also more accurate. “The new model assesses intersections by risk, not by crashes,” Slack-Smith explains. “Taking out the variability and analyzing by risk is a fundamental shift in how we look at this problem and make recommendations to reduce risk.”

While the project is in the early stages, a test scenario using the machine learning method estimated a 25% reduction in crashes compared to the previous method.

Easier sharing and more flexibility

The Commission opted to deploy the AI solution on the SAS Cloud, a strategic decision that provides numerous benefits, like expedited access to the software, the ability to easily share results with others and savings on infrastructure expense.

Greater self-sufficiency

Another key benefit of the solution is the self-sufficiency it provides. The flexibility of SAS Viya to serve analysts with various coding language preferences and proficiencies allowed Slack-Smith and his team to bring analytics in-house. This represented a quantum leap from old approaches, which often involved hiring transport consultancies or academics to produce reports and models – taking up to 12 months to receive.

The in-house capability was realized following a few weeks of training for analysts with varying degrees of SAS experience. “We've lowered the threshold to doing advanced analytics, and we can now do it ourselves,” Slack-Smith says.

Looking at its overall mission, the Commission’s use of powerful, fast and flexible AI has transformed how it fulfils its mission of promoting road safety. By combining the automated accuracy of AI with the nuanced judgment of human intelligence, the Commission is on course to profoundly improve the safety of Western Australia roads and save lives.