Resilience Through Algorithms: Forecasting Sustainable Development Amid Climate and Geopolitical Uncertainty

This study explores how climate vulnerability, geopolitical risk, financial development, economic complexity, and governance quality influence sustainable development across 17 low‐ and middle‐income countries (LMICs) during 1995–2024. Using machine learning, ensemble, and deep learning approaches, we compare predictive performance against traditional regression models to evaluate the Sustainable Development Index (SDI). The results show that while baseline linear models explain only about 38% of SDI variation ( R 2 ≈ 0.38), ensemble methods substantially improve predictive accuracy, with the Voting model achieving an R 2 of 0.817. Deep learning models perform moderately ( R 2 ≈ 0.58), remaining below ensemble approaches. Generalized Additive Model (GAM) analysis confirms statistically significant non‐linear effects ( p < 0.01) across predictors, and partial dependence analysis identifies a climate vulnerability threshold near 0.47, beyond which SDI declines more sharply. SHAP‐based interpretability consistently identifies climate vulnerability as the most decisive factor, followed by geopolitical risk and governance effectiveness, while financial development and economic complexity play comparatively weaker roles. These findings suggest that finance ministries, environmental agencies, and development partners should prioritize climate‐resilient infrastructure standards, adaptation financing mechanisms, and conflict‐sensitive investment planning, particularly in LMICs characterized by high climate vulnerability and institutional fragility.