Too often, academic efforts fail to demonstrate their real-world benefits. At the same time, many businesses and government agencies don’t adopt new technologies quickly enough to solve their pressing problems. In the case of MIT and the Dallas Fire and Rescue Department (DFR), however, a unique collaboration broke through typical logjams.
About two years ago, the department—which serves approximately 1.3 million residents—was struggling to inspect commercial properties for fire hazards in a timely, efficient manner. With 120 inspectors and 47,000 properties, they were resource-constrained and it was difficult to prioritize the workload.
Officials knew that they needed to sort through historical data about local buildings—such as their age, last inspection, or violations — that would point them to those with the highest probability of fires, according to Chief Ben Thornton, but the data wasn’t consistent or easy to correlate. Without “clean data” the department couldn’t prepare and prevent these fires. “Even if we couldn’t predict when a fire would happen, we could look at the likelihood of problems that lead to fire” if the data was available.
Fortunately, Dallas’ Chief Innovation Officer, Laila Alequresh, was researching ways to expand her team’s capacity and leverage university talent to support data analytics projects for the city at that time. When she learned about the DFR in 2019, she “put us in touch with the MIT A-Lab and we presented our needs,” Thornton said. “We are very grateful that she set this in motion.”
The MIT Analytics Lab (A-Lab) course challenges graduate student teams to solve a real-world problem using analytics, machine learning, or other digital technologies. Most of the problems are at for-profit companies, so the Dallas project was somewhat unique. The students work for nearly a year with sponsors to develop their models, refine their results, and then present their findings– first to the sponsor, and then to judges. Twenty-three MBA student teams reported on their research last December. A panel of academic and business experts chose one winning team and two runners-up, including DFR.
The Dallas-student team–Akhan Mukhanov, Jonathan Tukiman, Shenheng “Angela” Xu, Qijia “Nicole” Zou– was mentored by Nicholas Van Niel, an MIT Sloan alumnus and manager at Analysis Group in Boston. They worked with Thornton and others in DFR to develop a predictive algorithm trained on historical fire incident and inspection data as well as business permits and property condition information (the full team–students and sponsors– is pictured above).
They merged very large databases from 2007 to present using known address files, code violations reports, and other data that was sometimes hard to find or match up. Once that was done, they were able to identify high-risk areas and likelihood of fire by location, property type, and other factors.
“We ran into difficulties that the students could solve quickly. They plowed through it and produced the dataset,” Thornton said. It was “thrilling” to work with such accomplished students, he said, and “astounding” that they created the database in only six weeks.
“We had weekly call meetings, for clarifications and to offer our experiential knowledge,” he said, but “they did the hard work of data merging; we would never have been able to complete this in such a short time.”
Just as the Dallas team had to become familiar with terms like fuzzy matching algorithms and random forest modeling, the students learned that the benefits of their work went far beyond data analytics. “We showed them that they were saving lives– children, disadvantaged populations– and boosting firefighter safety,” Thornton said. “The program was making the community more livable and safe. They didn’t see the depth at first.”
Other fire departments around the country –including a project in Pittsburgh in conjunction with Carnegie Mellon University, as well as plans in New York and Portland, OR–are implementing similar techniques.
Keith Wilson, Senior Fire Inspection Officer at DFR, said the database “is changing how we do our job. Seven years ago, inspectors walked the streets to find violations, and they were not hitting problem areas. Now, our work is more targeted and we can identify risk factors and go after them.” Previously, the work seemed “overwhelming,” he said. Now, inspectors have higher morale because they see progress.
With the back-end technology in place and inspections under way, Innovation Officer Alequresh said a portal and dashboard for the department will be created. More broadly, Dallas’ new Mayor Eric Johnson and city leadership want more emphasis on data-based decision-making, she said, so “we are sharing these results broadly” and looking for other projects that demonstrate the power of data analytics in city government.
“We could not have gone out and contracted this work,” she noted, because of the high cost and time it would have taken. In addition to saving hundreds of thousands of dollars, “we’re saving lives,” she said, and you can’t put a price tag on that work.
