Carnegie Mellon University

Pittsburgh fire

April 04, 2018

City and Carnegie Mellon Highlight Fire Prediction Analytics Work

Mayor William Peduto joined officials from the Departments of Public Safety and Innovation & Performance and from Carnegie Mellon University to detail their latest collaborative efforts to use data sharing to provide Pittsburgh residents with better city services. 

The Mayor lauded a predictive model built by CMU Ph.D. student Michael A. Madaio that determines fire risk in the city's commercial buildings, which since last year the Fire Bureau has used to help inform its fire inspections. There are roughly 22,000 commercial properties in the city and the Bureau cannot inspect every last one, so the model from Madaio, a doctoral student at the Human-Computer Interaction Institute, helps the Bureau prioritize its inspection work. 

The model was built with data including historical fire incidents from 2009-2017, property assessment and valuation data and non-fire related inspection data from the Department of Permits, Licenses, and Inspections. It is updated every week. 

In the project's first six months, beginning in July 2017, the model identified 57 properties with a high-risk of fire. Of that number 50 of the properties (or 88%) indeed experienced fire incidents. 

"The technology being pioneered in Pittsburgh is making residents safer, and it is proven to work," Mayor Peduto said. 

New York City and Atlanta have similar predictive modeling technology for fires but only the model designed for Pittsburgh is updated every week, using assistance from Innovation & Performance and the Western Pennsylvania Regional Data Center

The Mayor said similar predictive analysis is planned for the Bureau of Police and that Pittsburgh – working with university and foundation partners – is poised to become a worldwide leader in the predictive analytics field. 

Further details on Madaio's work are available here.

Coverage of the event:

"Where's the fire? This data approach is helping Pittsburgh pinpoint buildings at risk" - Pittsburgh Post-Gazette
"50 of 57 buildings rated 'high risk' for fire in Pittsburgh called for help" - Pittsburgh Tribune-Review
"Data Tool Created By CMU Student Helps Predict Fire-Prone Buildings In Pittsburgh" - KDKA
"CMU program helps Pittsburgh predict where fires will break out" - WPXI
"AI Helps City Protect Buildings At High Risk For Fire" - 90.5 WESA