In the end, it was the team that went the extra mile—literally—that won the SAS Institute Safe Roads Competition.
After poring over aggregate accident data from the Toronto Police Service and fleet telematics company Geotab Inc., Corinne Blanchett and Farid Rasi put the results of their analytics assessment in context by visiting the sites of a number of fatal pedestrian accidents from 2018, either online or in person. This included what turned out to be a hair-raising effort to navigate an east-end crosswalk flagged by the data, and an in-depth in situ analysis of a fatal August accident in Scarborough.
“You clearly risked your lives for this,” said TPS data science lead Meghan Fotak, presenting the honours to Blanchett and Rasi.
Teams from George Brown, Humber, Seneca and Georgian Colleges competed for prizes of a SAS certification exam, a $180 USD value. Members of the top three teams collected the prizes, with Blanchett and Rasi also being treated to box seats to an upcoming Toronto Football Club match at BMO Field.
It wasn’t the crash-site visits that won the competition for the pair, though. “That was just the cherry on top,” said Geotab data scientist Dena Krieger. The visits “confirmed the level of care” that had gone into the analysis, said Fotak. But the strategy of drilling down into a specific problem and delivering a “clear narrative” that impressed judges most, Fotak said.
Teams worked with KSI (killed or seriously injured) incident report data from TPS, along with aggregated Geotab fleet telematics showing traffic speeds, acceleration and hard braking, road conditions, and more. Presentations explained the methodology used, the conclusions drawn and recommendations for TPS.
Blanchett and Rasi’s analysis flagged a particular KSI risk to pedestrians 55 and older in several clusters of Toronto wards. Focusing on that problem, the pair dissected a representative accident on the ground. Combined with the pair’s data crunching, the analysis revealed a number of common risk factors: presence of older populations and long stretches without a marked crossing were major contributors.
For its recommendations, the pair took a page from the originators of the Vision Zero Network, a strategy to eliminate traffic deaths that has swept Europe and North America since being implemented in Sweden in the late 1990s. “In every situation, a person might fail, the road system should not,” according to the Swedes. Blanchett and Rasi’s recommendations stressed self-enforcing road design features and controls—roundabouts, speed bumps, and the like—over awareness campaigns, which they felt are less effective.
“They realized that analytics works best when you get granular,” Krieger said.
While it’s the first road safety-specific contest, SAS has been running similar events for college and university students globally, said Mark Morreale, academic program lead for SAS Canada. “We want to invest in education,” to make connections with students, he said.
“We recognize there’s a keen need for data scientists,” he said.
Open data fosters educated public
May Masoud, SAS data sciences solutions specialist and the company’s representative on the judging panel, stressed that the students in the competition were not trained data scientists or modelers—they merely had access to the data sets and analytics tools required.
“This is the importance of easy-to-use technology,” she said.
Fotak said the point of the exercise was to explore and engage with the data to derive insights. Rather than propose a solution to a business problem, the competitors had to define one. Rather than a hypothesis-driven approach—given this problem, what data do we need to solve it?—students were to take a data-driven approach: Given this data, what can we learn?
“That is exactly what you do as data scientists,” Fotak said.
Through its Public Safety Open Data Portal, TPS is working with college and university students on an ongoing basis, Fotak said. Engaging the public with open data is part of the mission of the TPS Transformational Task Force. The force’s crime, traffic and boundary data—more than two dozen datasets in all—is freely available for crunching and analysis.
Geotab’s open data store draws information from 1.5 million customers North America-wide. For privacy purposes, data points are aggregated from multiple customers. Where there’s a significant issue, there’s no shortage of vehicle reports, Krieger says.
Geotab collects on the order of four billion data points a day—GPS and cellular data, telematic data like harsh braking, acceleration or turns, hyper-local environment data like temperature and pressure. It’s from a subset of overall traffic—Geotab’s customers drive commercial vehicles, so are “systematically different” from general traffic in terms of routes and access points—so it’s not definitive, but can be used to inform broad conclusions, Krieger said.
Among those conclusions for Toronto drivers: All of the presentations flagged eastern Scarborough and the region near Pearson International Airport in the west end as higher-risk areas for accidents. You’ve been warned.