The predictive modeling tool was created by a public-private team of coders and data scientists and funded by Bloomberg Philanthropies' Innovation Team program.
The Internet of Things and predictive data analytics are two of 2016's most intriguing technologies, because they foreshadow our vision of the future. The idea that lightweight sensors, placed anywhere, and the data they generate can create a more accurate understanding of the environment holds great promise for concepts that just decades ago were considered fantasy. Knowing where lead poisoning might occur next, where a bomb might go off, or which pipe is ready to burst marks a huge advance in data science.
When Bloomberg Philanthropies began its Innovation Teams program in 2011, the idea was to support new ideas, methodologies and technologies that government was interested in, but couldn't necessarily afford. The i-team initiative now has 20 participating cities, and has resulted in breakthrough work like New Orleans' drive to reduce murder through data analysis, and Centennial, Colo.'s mission to relieve roadway congestion via traffic pattern analysis.
One of the program's newest initiatives is a predictive model now being tested by Syracuse, N.Y., to identify which of the city's water mains may break next. In Syracuse, fixing water main breaks is a constant and expensive problem, with the city averaging almost one break per day. If the six-person team developing the predictive pipe system is successful, it will save Syracuse time and money while possibly making the technology popular with just about any city government.
The potential success of projects like these gives more credence to the notion that local government ought to orient more of its long-term thinking around technology and innovation. Even among technology's proponents, what the structure of those efforts should look like remains an ongoing debate. While public leaders frequently wonder how they'll pay for innovation with so many items already on the budget, the Bloomberg i-teams represent yet another avenue for experimentation.
Syracuse's team consists of five full-time employees and one part-time professor who lends research and GIS expertise. With the exception of the team's director, Andrew Maxwell, the members are new to government, hired on from the private and nonprofit sectors after Bloomberg's funding came through in spring 2015. Maxwell, who has worked for Syracuse government at the city and county level for the last 10 years, explained that bringing together a diverse range of nongovernment workers can generate interesting new ideas for the city's most challenging problems.
"Even though we're a mid-sized city, we experience hundreds of water main breaks a year," Maxwell said. "Decades of deferred maintenance, declining tax base, dwindling state and federal support — all these factors combine to create a crisis situation of sorts relative to city infrastructure, most notably, our water mains."
The Syracuse i-team was fully operational by May 2015, and around that time began cooperating with the Eric & Wendy Schmidt Data Science for Social Good Summer Fellowship program, out of the University of Chicago, to see what kind of predictive modeling might help the city if they shared the right data. The program is pointing its predictive powers at a range of societal challenges, including reducing the incidence of lead poisoning by partnering with the Chicago Department of Public Health. In Syracuse, work began when the i-team went hunting for data.
"First, it was primarily our team here in Syracuse working with the water department to identify all the data we could get our hands on," Maxwell said. "Luckily we had already spent several months doing that anyway, so we were pretty prepared for this partnership with the fellows. We worked together to compile that data in the most usable formats and passed that onto them. Additionally, the fellowship team did come here to Syracuse for a couple days and they had the opportunity to visit the water department themselves and get their hands on different data sets and begin to firsthand develop that modeling."
After months of gathering and sorting through data, the team now has a predictive model that contains a risk score for each city block. The risk score is based on metrics like soil conditions, pipe materials and break history. The model's efficacy will be tested in the coming months.
"We've begun trying to sink our teeth into what they've developed for us, and we're excited about the opportunity to begin taking that more proactive approach and developing that predictive model," Maxell said. "We're trying to understand how we can engineer it here and still working out some of the questions for the fellows that have developed it. We're not ready to begin releasing data on the risk scores, but we're feeling really good about the raw data we've seen so far as the group has finished up their work and handed in that deliverable."
Main break prediction is the i-team's focus now, but there are other components to its infrastructure monitoring work. A pilot program equips pipe valves with sensors that listen for small leaks. The sensors automatically populate a database with audio frequency data that the city can access via a Web interface and perform diagnostics to spot potential problems.
"[We're able to see] how large those problems might be with a great degree of accuracy, both in terms of the sound frequency and the location," Maxwell explained. "Then as we look at that data, we're able to send out water crews to go personally to those locations to sound the system and that's a little more manual — they listen, identify the location and we're able to do a much smaller, more surgical dig into the street to locate and fix the leak."
The team also runs a pilot, called SQUID, similar to the Boston Mayor's Office of New Urban Mechanics' Street Bump project. The Syracuse program uses images and accelerometer data accessed via a Raspberry Pi (a computer the size of a credit card) to rate the roughness of roads. That data is automatically sent to a database, which helps the city plan how it deploys its road maintenance crews more efficiently.
The future of a program like Syracuse's i-team is based on results, politics and a government's willingness to review its organizational structure and priorities. Syracuse's three-year Bloomberg grant expires at the end of 2017. The water main break model was designed to predict breaks through the end of 2018. By comparing historical data from 2016 with the model's predictions, the team will soon be able to validate its ability to see into the future and make adjustments as needed.
"Part of this process is creating a transition plan, essentially establishing a more sustainable pathway financially for the group and that can incorporate different strategies," said Maxwell. "We've budgeted funds through the city's annual general fund budgeting process [and looked for] grant funds or private funding. A big part of our work has been creating this innovation capacity, developing new initiatives that we believe are effective. It's really demonstrating the value of this sort of approach and by doing so I think we make a good case that this team should be a permanent part of city government and should be funded through the city budgeting process accordingly."
Making the team a permanent fixture has already happened to a small extent as one team member was named the city's first chief data officer. Beyond Syracuse, numerous cities across the country are ready to adopt successful programs developed by Bloomberg's i-teams. Syracuse is a great test case, Maxwell said, because the city is large enough to have big problems, but small enough to keep those problems manageable.
"We are of a size where we reflect a lot of the challenges people experience, including big cities," said Maxwell. "But we are small enough that a lot of these new initiatives and concepts can be prototyped, developed, measured and implemented much quicker. We work closely with department heads and senior officials who are responsible for these issues. And for those initiatives we find to be successful, we're well positioned to scale them up to a citywide level. I think that puts us in position to be a great test bed for how these innovation teams that Bloomberg has been developing can lead to new solutions that other communities can then learn from."