But how does the population of vehicles in an area relate to the local demographics? To find out, the team trained another deep-learning algorithm to learn the correlation between vehicle types and the data from U.S. Census and presidential election voting patterns in each precinct (an area of about 1,000 people). This training data set consisted of the data from 35 cities.
They then used the rest of the data to test the deep-learning algorithm. The question they wanted to answer was: given the pattern of vehicles in an area, could the algorithm accurately predict the demographics as recorded in the U.S. Census and presidential voting data?
It turns out that the deep-learning algorithm can do this remarkably well. “Using the classified motor vehicles in each neighborhood, we infer a wide range of demographic statistics, socioeconomic attributes, and political preferences of its residents,” they say.