Big data is sexy these days but still it’s a big deal to get coverage in the New York Times for research conducted in Rwanda. Josh’s work is complex and involves training data sets and also the use of multiple kinds of data.
He and his colleagues relied on anonymized data on billions of interactions, including details about when calls were made and received and the length of the calls. The researchers also looked at when text messages were sent, and which cellphone towers the texts and calls were routed through in order to get a rough idea of geographic location.
“So it’s the who, where and when of the call, but not the what or the why,” Dr. Blumenstock said.
They combined this information with responses collected from about 850 cellphone owners to build an algorithm that predicts how wealthy or impoverished a given cellphone user is.