Application of a deep learning algorithm for predicting socioeconomic data through satellite images in the Vale do Ribeira

Undegraduate thesis presented at the University of São Paulo. Abstract This work has the objective to train a machine learning algorithm, using satellite images and socioeconomic data. To predict socioeconomic indexes in the Vale do Ribeira’s territory. For that, the project was divided into two parts. The first one has the objective of reproducing the reference article Jean, N., et al. 2016, comprehending the used methods aiming it’s a reproduction on national territory. The second one, objectifies the appropriate adjustment to predict Brazilian human development indexes, starting from IBGE’s census data and satellite images of the Brazilian territory, specifically bounded to the Vale do Ribeira’s territory, for being circumscribed on many protected areas. To do so, the processes, charts, and algorithms from the first part will be used. Based on this approach it is expected to study the efficiency and accuracy inside Brazilian territory. All generated results and accuracy about the algorithm will be shared with the scientific community, more specifically the PARSEC project. The group expects to create algorithms of censitary data predictions, scientific divulgation, and worldwide data sharing.Acknowledgments: The PARSEC project is funded by the Belmont Forum, Collaborative Research Action on Science-Driven e-Infrastructures Innovation