From Traditional Models to AI: Enhancing Financial Credit Risk Assessment Through Machine Learning

This thesis explores the transition from traditional credit risk assessment models, such as Altman’s Z-score and logistic regression, to advanced artificial intelligence (AI) techniques. Using real-world loan application data from IDX Partners, the study evaluates and compares machine learning models including Logistic Regression, Random Forest, Support Vector Machine, and Neural Networks. The research addresses challenges of class imbalance, interpretability, and regulatory compliance while highlighting the importance of variables such as interest rate, loan amount, annual income, and debt-to-income ratio. Findings show that AI models enhance predictive accuracy and provide valuable tools for financial institutions, but model selection must balance performance, transparency, and business context.