Aug 7, 2017
In this episode we focus on using data to help predict where home fires may occur and technology to help detect them in vacant properties. This episode also ties into an earlier one Data and Smoke Alarms in New Orleans that you might be interested in. Using data to predict fires e start by talking with Jon Jay, a doctoral student at the Harvard School of Public Health, and author of the article "Can Algorithms Predict House Fires." In this article, and on this podcast, Jay talks about how it is possible to use data to predict where house fires are more likely to occur in a city so that the fire department can better focus its prevention and home safety visits, and the results are astounding. By randomly visiting neighborhoods, it was determined that a fire department would probably visit 20% of the homes likely to have a fire. However, by using a data-driven approach, this percentage jumped to 71% resulting in a much more effective and efficient use of staffing. The data that was used is from publicly available sources, and by bringing together a variety of data sources it is possible to refine the model significantly and really hone in on the geographic areas of the city as well as the different property types. Using technology to detect fires We then have a conversation with Nathan Armentrout from Eidolon who invented CASPER - Continuous Autonomous Solar Powered Event Reporter. This is a device that is installed in vacant homes and listens for the sound of a smoke alarm in the building then sends out an alert by cellular signal to the cloud. This device came about as a result of a hackathon sponsored by Louisville, Kentucky, where there was a problem of fires in vacant homes. These fires were creating significant risks to the other homes around them because they were not being detected until they were well developed instead of in their early stages. The city put the problem forward to the Level 1 hackerspace, and competition was put together to find a solution. Armentrout developed CASPER which has been successfully pilot tested in Louisville and he is now rolling out in beta to other cities.