Explainable AI–RegTech early warning systems for financial crisis detection: integrating governance and macroeconomic indicators for regulatory supervision

Purpose This study aims to develop an explainable artificial intelligence (AI)–regulatory technology (RegTech) early warning system (EWS) to predict economic and financial crises by integrating macroeconomic, financial, governance and labour-market indicators within a sustainability-aligned supervisory architecture linked to Sustainable Development Goals (SDGs) 8 and 16. Design/methodology/approach Using a harmonised panel of 30 countries (1980–2024), the authors construct a rare-event predictive framework combining logistic regression and ensemble learning under severe class imbalance. Model performance is evaluated using recall-sensitive metrics and policy-weighted loss functions. Shapley additive explanations-based explainability ensures regulatory transparency, and predicted probabilities are transformed into calibrated supervisory risk-alert bands. Findings While ensemble algorithms achieve high overall accuracy, they exhibit weak minority-class detection. Logistic regression demonstrates superior crisis recall (approximately 0.71), highlighting the importance of recall-oriented optimisation in supervisory contexts. Governance quality, inflation volatility, credit expansion and labour-market fragility emerge as dominant systemic risk drivers. Originality/value To the best of the authors’ knowledge, this study proposes the first SDG-aligned, explainable AI–RegTech crisis prediction architecture explicitly designed for supervisory deployment. By reframing crisis surveillance as a multidimensional governance challenge, the framework bridges predictive analytics, RegTech and sustainable development monitoring.