Description
In this Talk, we will discuss the process of analysing terabytes of GeoTIFF images of surface lights on Earth from space at scale using Multiprocessing, geospatial, image-processing and raster libraries on high-performance AWS instances. Second, we will discuss how the data can then be used with Machine Learning libraries to predict GDP and other economic metrics, especially during supply-demand shocks like COVID-19. In a recent analysis, using this approach, we accurately predicted the GDP impact in India during COVID-19 with a 99.5% accuracy, much ahead of official announcements, compared to 67% accuracy of estimates from leading investment banks. Overall, the Talk covers 4 disciplines: 1) computer science - TB-scale parallel processing using Python/Linux, 2) stats/ml - geospatial statistics & econometrics, 3) engineering - high-performance cloud computing and 4) finance - economics of supply-demand shocks.