Neural network forecasting algorithm as a tool for assessing human capital trends of the socio-economic system

Бесплатный доступ

The article addresses the issue of neural network forecasting of human capital size, structure, and dynamics. The object of the research is the socio-economic system. The subject of the study is the practice of applying neural network models to forecasting socio-economic indicators, human capital in particular. The purpose of this work is to apply neural network modeling and adapt its algorithms to build a forecast of the studied indicator for the future. The statistical base is data on demography, the investment volume in human capital of the regional economic system, as well as the indicators of socio-economic development. The investment volume in human capital is determined by budget and private citizens’ expenditures. To forecast the human capital dynamics the authors used the values of investment volumes the forecast of which, in turn, is built using a neural network modeling. The neural network model used in this study is a multi-layer fully connected perceptron with a sigmoid logistic activation function. Neural network modeling of forecast values of investment volumes has shown its effectiveness. Human capital assessment for the period of 2000-2018 and its forecast for the period of 2019-2023 are based on the example of the regional economic system of the Udmurt Republic. Our calculations show that the highest growth rate of the studied indicator has been demonstrated since 2013, and its further increase is predicted. The results obtained correlate qualitatively with the dynamics of changes in the Russian human development index, determined by the UN experts. The proposed method of calculating and forecasting human capital can be used to assess and compare the socio-economic situation of Russia’s regions.

Еще

Human capital, neural network modeling, algorithm forecast, investments, socio-economic system

Короткий адрес: https://sciup.org/147225290

IDR: 147225290   |   DOI: 10.15838/esc.2020.6.72.7

Статья научная