إطار تنبؤي للناتج المحلي باستخدام نموذجي Ridge-ARDL ونموذج Random Forest خلال الفترة 1960-2023

Abstract

This study develops a hybrid forecasting framework for Syria’s gross domestic product (GDP) by integrating a Ridge-regularized Autoregressive Distributed Lag (Ridge-ARDL) model with a Random Forest algorithm. The research addresses the challenges posed by high-dimensional macroeconomic data and complex interdependencies among variables. The framework begins with constructing the Ridge-ARDL model to estimate long-run relationships and short-run dynamics, achieving strong explanatory power with an R² of 0.74 and a stable error-correction mechanism. The resulting linear predictions are then incorporated into a Random Forest model to capture nonlinear interactions and improve predictive performance. The hybrid model outperforms its individual components, reaching an R² of 0.85 and achieving a substantial reduction in RMSE. Variable-importance analysis reveals that knowledge-based indicators, such as the number of researchers and public expenditure on education, are key determinants of GDP in Syria. Using annual data from 1960 to 2023 and forecasting through 2029, the results indicate modest but positive future growth accompanied by wide uncertainty bounds, highlighting the need for structural reforms. The proposed framework provides a robust quantitative tool for policymakers, combining economic interpretability with the flexibility of machine learning to support more accurate and informed decision-making