FORECASTING OF SUSTAINABLE DEVELOPMENT INDICATORS BASED ON DIGITALIZATION TOOLS

Abstract

In today's conditions, the economic processes in Ukraine are significantly influenced by numerous factors, including the military situation, inflation, logistical difficulties, environmental problems, social inequality, the depression of industrial regions, and mass emigration. Migration processes, particularly those caused by the war, have had a particularly profound impact, leading to a decrease in human capital, labor shortages, demographic changes, and a reduction in the gross regional product in the affected regions. To ensure sustainable development, it is crucial to focus on preserving human capital and implementing digital technologies. Among the modern methods of time series forecasting, which allow for the analysis of economic indicators in the short- and medium-term, two main approaches are highlighted: classical methods, such as ARIMA and GARCH, and methods of data mining, including neural networks (ANN), support vector machines (SVM), k-nearest neighbors (kNN), and random forests (Random Forest). These approaches are commonly implemented using programming environments such as RStudio, MATLAB, and Excel with add-ons. However, for more convenient and cost-effective analysis, cloud platforms like Google Colab and Kaggle Notebooks are increasingly used, as they are free and do not require installation. An assessment of sustainable development indicators based on scientific research from 2018–2023, using SciVal and Scopus, shows a growing interest in the digitalization of regional development. Practical analysis of the dynamics of sustainable development in the Zaporizhzhia region revealed indicators such as population size, digital infrastructure development, industrial production volumes, and gross regional product (GRP). Forecasting these indicators using the GARCH, Random Forest, and ARIMA models highlighted their strengths. The GARCH model demonstrated stable growth with a gradual trend. Random Forest showed significant sensitivity to fluctuations, predicting both sharp increases in certain indicators (K5, K6) and declines. ARIMA provided more conservative forecasts, ensuring stability for a number of indicators. The choice of forecasting model depends on the goals: ARIMA is suitable for obtaining stable long-term predictions, while Random Forest and GARCH are more appropriate for accounting for dynamic changes. Cloud platforms significantly simplify this process, making it accessible to a wider range of users. Therefore, the implementation of digital technologies is a key element in ensuring sustainable development in Ukraine's regions.