ARTIFICIAL INTELLIGENCE IN ECONOMIC ANALYSIS AND FORECASTING OF THE REPUBLIC OF MOLDOVA

Abstract

The relevance of this study is driven by the growing need for high-precision macroeconomic forecasting tools for the Republic of Moldova — a small open economy that is historically highly exposed to external price shocks, significant energy import dependence, and continuous real sector volatility. Traditional forecasting methods have shown their limitations during recent crisis periods, emphasizing the demand for more advanced analytical frameworks. The purpose of this research is to conduct a comprehensive comparative assessment of the predictive capacity of three distinct model classes — multiple linear regression (MLR), Autoregressive Integrated Moving Average (ARIMA), and Long Short-Term Memory (LSTM) recurrent neural networks. These models are applied to forecasting Moldova's inflation rates and Gross Domestic Product (GDP) growth based on macroeconomic data spanning the period from 2018 to 2025. The empirical information base is derived exclusively from official publications provided by the National Bank of Moldova (NBM) and the National Bureau of Statistics of the Republic of Moldova. To ensure methodological rigor, the quality and accuracy of the models were evaluated and cross-validated using standard statistical metrics, specifically Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The comparative results compellingly show that the deep learning LSTM model demonstrates significant and consistent superiority over traditional econometrics. Specifically, it reduces the RMSE of inflation forecasts from 1.139 (linear regression) and 0.973 (ARIMA) down to 0.354, while improving the MAPE from 21.4% and 17.3% to a highly accurate 6.85%, respectively. Furthermore, a generated scenario forecast for the upcoming 2026–2027 period indicates that inflation will likely consolidate near 5.0% and 4.5%, which remains entirely consistent with the NBM's established target corridor. Additionally, GDP growth in the baseline scenario is projected at stable rates of 2.2% in 2026 and 3.0% in 2027. The practical value and novelty of this study lie in the potential direct integration of the proposed AI-driven predictive models into the strategic analytical systems of the National Bank and the Ministry of Economic Development and Digitalization of the Republic of Moldova, thereby enhancing policy formulation under uncertainty.