FORECASTING THE DYNAMICS OF DEMAND AND SUPPLY IN THE UKRAINIAN LABOR MARKET

Abstract

The research is aimed at determining the trend of changes in demand and supply in the labor market of Ukraine under martial law and forming a forecast of their dynamics using methods of time series analysis and artificial intelligence. A time series is a set of observations that were collected at fixed intervals of time. An example of a time series can be the number of resumes and vacancies in the labor market each month. Currently, there are many methods for analyzing and forecasting time series: autoregressive models, moving averages, their extensions, such as ARMA, ARIMA, and exponential smoothing. However, they have a number of shortcomings that can hinder forecasting. In particular, most of them assume linear relationships between observations which makes it impossible to identify and model complex long-term dependencies. To overcome these limitations, recurrent neural networks, in particular, their variety – long short-term memory (LSTM) – is actively used. Due to their architecture, LSTM networks can effectively learn nonlinear relationships and long-term dependencies in data. During the analysis of the source data, which reflects the dynamics of monthly changes in demand and supply in the labor market in Ukraine from January 2022 to February 2025, it was determined that there is no obvious seasonality, and recent observations have a more significant impact on current ones. The resulting model, which is implemented as a shallow LSTM network, showed low error values ​​on both data sets when assessing the quality of forecasting, signifying the correctness of its work and good generalization ability. The projected dynamics of labor demand demonstrate stabilization in the first half of 2025, however, the estimated level of its supply shows a gradual decline during this period. The forecast results indicate that the imbalance between demand and supply in the Ukrainian labor market will remain significant.