Bridging Temporal Dynamics and Nonlinear Patterns inMacroeconomic Forecasting: A Hybrid Statistical Learning Approach
Year: 2026
Type: article
Abstract: Accurate macroeconomic forecasting is essential for effective policy formulation, financial planning, and economic stability, yet it remains challenging due to structural changes, economic shocks, and complex nonlinear interactions among economic indicators. Traditional time series models often struggle to capture such complexities, while standalone machine learning approaches may overlook temporal dependencies. To address these limitations, this study proposes a Hybrid Predictive Framework for Macroeconomic Forecasting (HPFMF) that integrates complementary strengths of time series and machine learning models. Using annual macroeconomic data for more than 200 countries spanning 2010–2025 from the World Bank Open Data API, the framework applies systematic preprocessing, including missing value handling, winsorization, logarithmic transformation, and feature scaling. A time series hybrid combining ARIMA, Prophet, and Exponential Smoothing captures temporal dynamics and structural shifts, while a stacked machine learning ensemble of Random Forest, XGBoost, and Support Vector Regression models nonlinear interdependencies. These layers are integrated through validation-based weighting to generate robust forecasts. Empirical results show that the proposed framework achieves superior performance, with significant reductions in forecasting errors and an R² of 0.92. Country-wise and temporal validations confirm strong generalizability, demonstrating the framework’s effectiveness for reliable, policy-oriented macroeconomic forecasting.
DOI: 10.14419/3e55tj72
Institution: Amazon (United States)
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: bronze
OpenAlex ID: W7124279299
FWCI:
Cites:
Cited by:
Related to: