Machine learning for credit risk management
Details
- Author: Anton Kalinichenko
- Title: Machine learning for credit risk management
- Supervisor: Prof. Dr. Jörg Osterrieder
- Degree: Bachelor of Science
- University: Bern University of Applied Sciences
- Year: 2024
- Status: Submitted Paper
Summary
Credit risk management is crucial for financial institutions to accurately assess borrowers' creditworthiness. This thesis investigates the most effective machine learning (ML) models for credit risk prediction and explores the challenges and limitations of their implementation. Through a comprehensive literature review and practical experiments using the German Credit Risk and Indicators of Heart Disease datasets, the study evaluates models including Logistic Regression, Bayesian models, AdaBoost, Support Vector Machines, Random Forests, k-Nearest Neighbors, and Multi-Layer Perceptron. Key findings indicate that Logistic Regression, Bayesian models, and Random Forests are highly effective, with specific models performing best on different datasets. The study highlights challenges such as data quality, model interpretability, computational demands, and regulatory compliance. This thesis underscores the importance of balancing traditional and advanced ML models, data preprocessing, and model tuning, offering valuable insights for future research and practical applications in credit risk management.