İşletme Fakültesi Brown Bag Semineri | Finans |Dr. Öğretim Üyesi Levent Güntay

İşletme Fakültesi Brown Bag Semineri | Finans |Dr. Öğretim Üyesi Levent Güntay

Tarih: 29.03.2024
Saat: 13:00-14:30
Yer: AB2 345

Konuşmacı: Dr. Öğretim Üyesi Levent Güntay- Özyeğin University

Title: Machine Learning in Credit Decision-Making: Balancing Predictive Power with Explainability

Abstract: Ensemble classification algorithms in Machine Learning, like Random Forest and Gradient Boosting Machines, excel at precisely evaluating a borrower's credit risk. Yet, the opacity of these models often leaves decision-makers, customers, stakeholders, and auditors in the dark about how specific factors influence credit scores. This lack of transparency in black-box credit scoring methods can undermine trust in the banking sector, with potential adverse effects on business financing and growth. Addressing recent regulations from the European Union and the United States for transparent and explainable credit decision processes, we propose a novel surrogate modeling framework. This framework is designed to convert a non-transparent black-box credit scoring model into an explainable credit scoring model. The resulting surrogate model produces a transparent decision tree that closely matches the in-sample predictions of the black-box XGBoost model. Our findings indicate that the Surrogate decision tree model not only surpasses the benchmark decision tree in terms of performance but also closely approaches the accuracy of the most accurate XGBoost model. Overall, our findings demonstrate the feasibility of developing explainable credit scoring models with minimal loss in accuracy.