Improving Loan Classification Models to Strengthen the Financial Stability and Profitability of Commercial Banks

Loan classification models are essential for identifying credit deterioration and supporting sound banking decisions. This thesis examines an integrated approach to improving classification models through financial, behavioral, facility, collateral, sectoral, and macroeconomic information. The proposed framework combines probability-of-default estimation with calibrated risk thresholds and explicit managerial responses. The study emphasizes that predictive accuracy alone is insufficient: a useful model must also be calibrated, stable, stress-resistant, and economically valuable. Improved classification can reduce credit losses, strengthen provisioning, improve risk-based pricing, allocate capital more efficiently, and support financial stability. The thesis presents a conceptual model, key classification variables, evaluation criteria, and implementation recommendations for commercial banks.