Evaluating economic and market risk levels for proactive identification of systemic threats and improved financial resilience

Continuous risk evaluation is essential for understanding economic stability and long-term sustainability. Financial systems evolve due to market fluctuations, geopolitical events, regulations, and technological advances, making early detection of risks challenging. Unaddressed risks may lead to crises or loss of trust, emphasizing the need for efficient assessment tools. This study modeled Risk Level prediction as a supervised classification task using 11 input features and 3,243 samples. Data validity was confirmed through Isolation Forest, with no deletions, and evaluated using 5-fold cross-validation. Two model families were applied: Extra Trees Classifier (ETC) and Categorical Boosting Classifier (CATC). Hyperparameter tuning employed the Catch Fish Optimization Algorithm (CFOA) and Dollmaker Optimization Algorithm (DOA). Model performance was measured using Accuracy, Precision, Recall, F1-score, and AUC. Copula-based sensitivity analysis identified credit_score and stock_index_change as key risk factors. Among hybrids, CACF achieved the best performance (Accuracy = 0.9676, AUC = 0.9757), while ETDO recorded the lowest (Accuracy = 0.9306, AUC = 0.9479). The proposed pipeline accurately classifies risk levels and reveals actionable drivers. Overall, the framework enhances prediction accuracy and speed, enabling quicker identification of high-risk cases and supporting informed, reliable financial decision-making—thereby improving stability and sustainability across dynamic economic environments.