Abstract: This study investigates the forecasting of Saudi Arabia’s non-oil GDP by applying a set of advanced hybrid time-series models, in conformity with the strategic aspirations of Vision 2030 to stimulate economic diversification and minimize dependency on oil income. Using annual data spanning from 1990 to 2021, the research develops and compares eleven hybrid models that integrate classical statistical frameworks (such as ARIMA and ETS) with modern forecasting techniques, including neural networks (NNAR) and composite models (TBATS). Among these, four hybrid configurations(A-N),(N-T), (E-N), and the composite (A-N-T)—demonstrated greater forecasting ability, as shown by reduced error values across MAPE, RMSE, and MAE measures. The usage of hybrid models is theoretically supported by their proven capabilities to tolerate nonlinear dynamics, structural alterations, and patterns typically ignored by classic linear models. This methodological flexibility is particularly significant within the context of Saudi Arabia’s fast shifting economy, which is marked by substantial structural reforms under Vision 2030. The forecasting outcomes imply a continued rising trend in non-oil GDP through 2030, with expected values ranging between 2,054,978 and 2,163,028 million SAR. Further, the inclusion of 80% and 95% confidence intervals increases the dependability and practical usefulness of the estimates. The study contributes to the burgeoning literature on economic forecasting in developing nations by highlighting the efficiency of hybrid models in capturing complex macroeconomic patterns. It gives a data-driven basis for policymakers to assist strategic planning, assess reform progress, and enhance non-oil economic sectors in pursuit of long-term sustainability.

Keywords: Non-Oil-GDP, Forecasting, Hybrid Models, ARIMA, NNAR, TBATS, ETS

Pages: 92-107