Forecasting Regional Inequality from Time Series Models to Artificial Intelligence-Based Frameworks

This research investigates the performance of the traditional econometric models against the advanced AI-based approaches in forecasting regional inequality in Morocco. Recognizing that regional inequalities evolve in complex, non-linear ways, the research evaluates models for their predictive accuracy and practical applicability in guiding policy. Conventional methods demonstrate reasonable performance, particularly in capturing linear trends. Among them, the AR model achieved the lowest MAPE and RMSE. However, AI-based models—especially LSTM and XGBoost—outperformed traditional techniques. LSTM proved highly effective in capturing both short- and long-term behavior, while XGBoost offered strong accuracy with efficient runtime. Conversely, Support Vector Regression (SVR) underperformed, reflecting its limitations in handling complex patterns. The findings highlight the growing inadequacy of traditional models for forecasting modern inequality dynamics and suggest that advanced AI tools, particularly LSTM, can provide more nuanced and reliable forecasts. These insights support the integration of AI in regional policy planning.