Machine Learning Forecasting of Russian Economy, Energy Cooperation and Crypto Asset Fair Value
Currently, the main tool for predicting macroeconomic indicators is an econometric tool that has proven itself well in the exact Sciences. However, the time series of socio-economic processes are often characterized by constant volatility, levels and directions of relationships between macroeconomic indicators, which complicates the process of justification and development of econometric forecast models, including in terms of evaluating the model parameters and the reliability of confidence intervals of the obtained forecasts.
In addition, one of the most important tasks in modeling macroeconomic processes is the procedure for specification of the regression equation. However, this procedure is not unified, which leads to a variety of methods for carrying out the specification and criteria for model optimality.
Thus, taking into account the above, we can say that the existing methods of forecasting and modeling of macroeconomic processes based on regression analysis do not fully solve the problems considered and can be improved in order to increase the accuracy of forecasting and reliability of confidence intervals. As a result, research aimed at developing new approaches to regression analysis that offer a sustainable solution to existing problems in modeling macroeconomic processes seems to be quite relevant and can be used by Central banks and the country's leadership in making management decisions.
Thus, the research paper proposes to propose approaches to solving the above-mentioned problems of regression analysis and modeling of macroeconomic processes in order to improve the effectiveness of the resulting models.