Application of Artificial Intelligence Models in Predicting Oil Price Volatility and Its Impact on Nigeria's Economic Growth

Abstract

Oil price volatility constitutes the primary source of macroeconomic instability in Nigeria, a petroleum-dependent economy where crude oil accounts for 87% of export earnings and 68% of government revenue. Conventional econometric models – ARIMA, GARCH, and VAR have demonstrated persistent limitations in capturing the nonlinear, structurally asymmetric dynamics of contemporary global oil markets. This study develops, benchmarks, and validates five AI-based forecasting models: LSTM-GARCH hybrid, Bidirectional LSTM, Temporal Fusion Transformer (TFT), XGBoost with SHAP interpretability, and a stacked ensemble using daily oil price and macro-financial data spanning 2000–2024 (N = 6,008 observations). The stacked ensemble achieved the lowest RMSE (1.51), MAPE (2.63%), and highest directional accuracy (83.7%), significantly outperforming all econometric baselines (Diebold-Mariano p < 0.001). ARDL bounds testing confirms long-run cointegration between oil price volatility and Nigeria's GDP growth, exchange rate, inflation, and government revenue. Long-run ARDL estimates reveal that a one standard deviation increase in oil price volatility reduces Nigeria's real GDP growth by 0.342 percentage points (p < 0.001) while raising exchange rate depreciation by 0.584 percentage points and inflation by 0.463 percentage points. Time-varying parameter VAR (TVP-VAR) analysis demonstrates that the macroeconomic impact of oil price shocks is asymmetric negative shocks generate contractions 1.37 to 1.81 times larger than expansions generated by equivalent positive shocks and that the transmission intensity has strengthened post-2015. These findings carry profound implications for Nigeria's fiscal planning, monetary policy, and sovereign wealth management, and establish AI-based price forecasting as an essential tool for evidence-based economic governance in oil-dependent developing economies.