Reviewing Deep Learning Methods in the Applied Problems of Economic Monitoring Based on Geospatial Data

Development of modern observation technologies, increase of the amount of open data, and development of new approaches to their processing open new opportunities in carrying out applied research in the economic activity of people. The central approach in this field is the use of the deep learning methods in the data processing and analysis of their time series. In this paper, we review the basic (in terms of geospatial analysis) sections of deep learning, namely, increasing the resolution of graphical data, using transfer learning for optimization of learning processes, scaling deep neural network models, and analyzing time series using recurrent neural networks.