Satellite images show ratios of graduates
With the help of neural networks, the experts at the University of Applied Sciences in Kufstein used the city of Vienna as an example of the concentration of academics for their research in collaboration with the University of Applied Sciences St. Pölten. The concentration of graduates in certain regions gives information about things such as the quality of life in regions, urban areas, and smaller neighborhoods (Glaeser, Kolko & Saiz, 2001; Jauhiainen, 2005).
Using square grids that were subdivided into five categories, the concentration of academics in Vienna was able to be predicted by means of a self-learning algorithm. In the process, the researchers found out that the satellite images contained certain patterns that correlated with the density of people who have a university degree.
In an initial attempt, the network correctly predicted 40.5 percent of the 3,314 grid cells (Koch, Dispotovic, Thaler, et al., 2021). This corresponds to a 100 percent increase of precision in comparison to a random classification of categories (20 percent probability for five categories).
The results from the first study show a high potential for neural networks and machine learning for urban and regional economies. In particular, it should be possible to predict demographic key figures in metropolitan areas with low amounts of available data. The determination of the number of academics also only served as an initial demographic parameter. In the future, other parameters should be verified to see how predictable they are. This could make it possible to optimize urban planning and development and could be more extensively controlled.
University of Applied Sciences Professor Dr. David Koch, Dr. Miroslav Despotovic, MA, and Simon Thaler, MSc, from the Institute for Energy, Facility & Real Estate Management at the University of Applied Sciences Kufstein Tirol. In cooperation with University of Applied Sciences Professor Dr. Matthias Zeppelzauer from the University of Applied Sciences St. Pölten.