Researchers at Stanford University have designed a probe using predictive analytics to observe poverty from space. Published in the journal Science, it uses an algorithmic model contrary to present methods that source data from old and out-of-date surveys. From space, a particular region’s or country’s brightness determines how developed it is; the twinkling lights, however, don’t inform which neighborhoods within them are destitute.
Other methods such as dispersing study groups throughout targeted areas would be overly time-consuming and expensive, so scientists are using computers to interact with satellites. They need to use, nonetheless, a particular approach. Fundamentally, researchers need a set of data that tracks where people live, and data from the public domain is the most convenient data trove.
A smaller more specific cross-section of data is then used before dropping all the information into a computer that can sift out reliable references. Stanford analysts applied their model to five African countries including Nigeria, Tanzania, Uganda, Malawi, and Rwanda. They first used nighttime images caught by the U.S. Air Force Defense Meteorological Satellite Program. Areas that were recorded as economically developed were compared to darker counterparts that were not.
The nighttime images were juxtaposed to higher-resolution daytime images available through Google Static Maps. The program was able to identify certain shapes within the wealthier regions that signified economic development.
“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,” said study lead author Neal Jean, a computer science graduate student at Stanford’s School of Engineering.
Statistical methods were then applied to particular items cited in the daytime and their correlation to income data culled in the surveys. Type of roofing material, for instance, was directly linked to income and a location’s distance from an urban center. The predictive analytics component of the program measured two crucial determinants of poverty: average spending by households and average household income.
Rwanda, for example, was referenced for the study’s success. The model predicted average household income more accurately than data collected from cellphones, according to the study. The model was effective in analyzing data and making predictions from other countries as well. Analysts noted that it was a resourceful tool writing that the program “is straightforward and nearly costless to scale across countries.”
In an essay that supports the observation, Joshua Blumenstock of UC Berkeley’s Data Science and Analytics Lab believes the surveyed data can “make it possible to differentiate between poor and ultra-poor regions,” which, in turn, “can help to ensure that resources get to those with the greatest need.”
Source: LA Times