In this project, credit scoring methods that are used in literature for credit risk assessment are taken into account. These methods do have great importance for lending institutions such as banks. Based on real life realizations, for credit scoring at first we aim to determine the significance of each variable in credit scoring. We test different combination of them to increase the performance of methods. Our goal is to compare credit risk scoring methods performances objectively and eventually suggest a tool for banks to decrease judgement bias and errors made by credit specialist which will decrease the default rate of consumer credits which will increase the banks' profits. Machine learning classification methods are implemented to reach proposed goal. The main performance comparison methods in the literature are used to evaluate each method's performance as objectively as possible.

Collaborators

• A. Sevtap Selçuk- Kestel,  Institute of Applied Mathematics, METU (Director)
• Ömür Uğur, Institute of Applied Mathematics, METU  (Researcher)
• Mervan Aksu, Institute of Applied Mathematics, METU  (Researcher)
• Oğuz Koç, Institute of Applied Mathematics, METU  (Researcher)

Funded by  GAP-705-2018-2780, May 11 2018 -  May 11 2019