Using Foursquare Data to Redefine a Neighborhood
Carnegie Mellon researchers believe they can capture the essence of an area based on what Foursquare users do there.
Defining the makeup of a particular neighborhood can be tricky. Locals may agree on the general area and character of, say, Manhattan’s Upper West Side, but we all have different opinions about what really goes on there, or even what its precise boundaries are.
With this in mind, a research project called Livehoods, from Carnegie Mellon University’s School of Computer Science, aims to shed some light on how people really inhabit their cities—and how this changes over time—by mapping data collected from 18 million Foursquare check-ins that have been sent out via Twitter.
The project groups check-ins by physical proximity and measures “social proximity” by how often different people check into similar kinds of places. The resulting areas—which may not correspond to what the locals typically think of as their neighborhood—are dubbed “livehoods.” As more people use smart phones and services that log their location data, interest in mining this information is likely to grow—both for the cultural insights it could show and the possibilities for advertising and other moneymaking opportunities.
Currently, Livehoods offers maps of just three places—Pittsburgh, the San Francisco Bay Area, and New York City (plus a bit of New Jersey)—on its website. The group hopes to add more cities soon, and says visitors can vote for ones they’d like to see mapped out.
Norman Sadeh, a computer science professor at Carnegie Mellon who is working on Livehoods, says social media can help define an urban space’s characteristics because it “really speaks at such a finer level than the data people have been relying on in the past,” such as census data.
Looking at the Livehoods map of New York is a little confusing at first, with clusters of colored dots all over the place, each hue indicating a specific livehood. You can click on any dot in a cluster to see the boundaries that the Livehoods algorithm has determined for that livehood, as well as the check-in location that corresponds with that particular dot.
Rather than taking on existing neighborhood names (which wouldn’t work since, in many cases, a single livehood can stretch into more than one neighborhood), each livehood has a number, as well as details about the most popular destinations and activities in that livehood. There’s also data about when people most often check in and where (not shockingly, lots of folks check in on weekends, and restaurant check-ins are among the most popular). Eventually, Livehoods may crowdsource the names of the livehoods, says Jason Hong, a research team member and associate professor at Carnegie Mellon.
Crunching all this data also means the Livehoods team can determine the most related livehoods—that is, the areas that are most visited by the same people.
“It’s a really interesting way to see a snapshot of the structure of the city,” says Livehoods team member and Carnegie Mellon graduate student Justin Cranshaw.
Sometimes there are surprises, too. I live in Livehood #44 in San Francisco, which covers much more ground than what I usually think of as my neighborhood, near the famous intersection of Haight and Ashbury. And the neighborhoods most related to mine aren’t exactly what I figured, either.
Beyond just being a resource for the curious, Livehoods can give residents better insight and understanding about their city, the researchers say. Sadeh suggests the data could be used to help stores determine where their customers are really coming from so they know where to advertise, or to make predictions about how changes—the addition of a Whole Foods supermarket, for example—may impact a neighborhood.
Already, urban planners in Pittsburgh have indicated an interest in working with Livehoods, Sadeh says.
“They are clearly seeing lots of different possible applications for this analysis,” he says.