Explainable Artificial Intelligence for Inflation Forecasting with SHAP, Random Forest, and LSTM: An Application to Algeria

Accurate inflation forecasting is crucial for effective policymaking, yet inflation dynamics often lack transparency. This study applies Explainable Artificial Intelligence (XAI) to analyze inflation in Algeria by combining SHapley Additive exPlanations (SHAP) with Random Forest (RF) and Long Short-Term Memory (LSTM) models. While LSTM better captures extreme inflation episodes and RF provides smoother forecasts, the main contribution lies in explaining model predictions. SHAP results identify food inflation as the dominant driver, followed by lagged producer inflation and the GDP deflator, revealing strong nonlinear and temporal effects, whereas energy inflation plays a limited role. In addition, the analysis highlights how machine-learning models can complement traditional econometric approaches by capturing complex interactions and regime-dependent behaviors that are difficult to observe with linear frameworks. Overall, integrating ML models with XAI enhances transparency, supports informed policy decisions, and provides robust, interpretable evidence to better understand and manage inflationary pressures in Algeria’s evolving macroeconomic environment context effectively.