Naturally, it’s not difficult for those with access to a satellite, or in this case common sense, to see that North Korea is a ridiculously shambolic economic mess, especially were you to look at the light levels compared to its considerably more developed neighbors. But researchers at Stanford have gone considerably further than that with algorithms and machine learning in the hopes of directing economic aid at areas within a country that truly need it.
Researchers take a novel approach to track poverty from space
Obviously, simply looking at light levels from space isn’t going to do the trick when it comes to those living in dire circumstances and abject poverty. It doesn’t matter how much you zoom you can call up from space you need considerably more.
And that’s where researchers from Stanford University came in according to a study published this week in the journal Science. While you can send an army of survey takers and census type investigators into perceived or actual poorer neighborhoods and regions; this is time consuming, involves no small amount of tens of millions and in certain areas nearly physically impossible in what can only be called the big world we live in at the end of the day.
Marathon Partners Equity Management, the equity long/short hedge fund founded in 1997, added 8.03% in the second quarter of 2021. Q2 2021 hedge fund letters, conferences and more According to a copy of the hedge fund's second-quarter investor update, which ValueWalk has been able to review, the firm returned 3.24% net in April, 0.12% in Read More
The researchers began their work with data in the public domain that showed where everyone in the world lived. From there the researchers required a considerably smaller data set that was known to be quite accurate. From there, the researchers fed a really powerful computer both sets of data and set it to work to overlay the two and to correlate nighttime light levels to regions known to be poor in Nigeria, Tanzania, Uganda, Malawi and Rwanda.
Using machine learning and Google Static Maps, the computer in question was able to use the nighttime data in conjunction with the static maps to determine areas of development and areas living in poverty.
“Without being told what to look for, our machine learning algorithm learned to pick out of the imagery many things that are easily recognizable to humans — things like roads, urban areas and farmland,” study lead author Neal Jean, a computer science graduate student at Stanford’s School of Engineering, said following the publication of the study.
“The idea is that if we train our models right, they help us predict poverty in areas where we don’t have the surveys,” he says, “which will help out aid organizations that are working on this issue.”
“In Africa, a lot of these places that are the most poor are actually just uniformly dark at night,” says Jean. “So if you use nighttime lights only to try to find these people, since there’s no variation in nighttime lights you can’t predict any variation in poverty.” Daytime imagery creates a fuller picture.
As you would expect from machine learning, the machine learns so the final computer model used to look at Rwanda was considerably more accurate in identify areas of abject poverty. That said, applying the same model to Rwanda and Nigeria didn’t work.
In an accompanying piece also published in Science, Joshua Blumenstock of UC Berkeley’s Data Science and Analytics Lab points out that the daylight data “makes it possible to differentiate between poor and ultrapoor regions.” This, in turn, “can help to ensure that resources get to those with the greatest need.”
What is transfer learning?
In order to get their computer to accurately track poverty and help organizations to direct aid in different countries requires something called “transfer learning.” It should also be noted that the researchers are well-aware that their method works considerably better in rural areas, villages and small towns than it does in cities. If you think of the difference between someone living in a penthouse in Central Park West compared to someone living on 223rd street in the Bronx, it’s easy to understand. The former could be making tens of millions a year while the other is making minimum wage but they both live in New York City.
In order to explain transfer learning, Jean surprised many with a film starring John Candy when he coached the Jamaican bobsled team to compete in the Calgary Olympics in Cool Runnings.
“The movie is about a Jamaican team that wants to compete in the bobsled event, but they don’t have ice, so they train for the Olympics by practicing on a dirt hill,” says Jean. “Dirt isn’t the same, but they learn transferable skills that they can apply to the actual task, which is competing on ice in a real bobsled.”
The team is quick to point out that looking at the African nations that it did is considerably easier than if it tried to map poverty in China or India. That said, the team believes it can be done and given the use of public domain statistics and imagery it can be done on the cheap.