Neural network modeling of the economic and social development trajectory transformation due to quarantine restrictions during COVID-19

Abstract

The article uses neural networks to model the effects of quarantine restrictions on the most important indicators of the country's socio-economic development. The authors selected the most relevant indicators and formed a specific sequence of its calculation to study the direction of transforming the trajectory of socio-economic development of a particular country due to quarantine restrictions. They used a multilayer MLP perceptron and networks based on radial basis functions. They chose BFGS and RBFT algorithms in neural network modeling. Collinearity study was the basis for data mining in terms of key factors of change. The author's approach is unique due to an iterative procedure of numerical optimization and quasi-Newton methods ("self-learning" and step-by-step "improvement" of neural networks). The model projected gross domestic product and the number of unemployed in the country affected by the COVID-19 pandemic over the three years.