Credit Risk Analysis with Classification Restricted Boltzmann Machine

Abstract

Banks are closely monitored by supervisory institutions because of the size and importance of the banking sector. One of the most important risks affecting the banking sector is the concept of credit risk. The banks try to make sure that the customers are well-financed and they can repay the given loans. But nowadays, as the number of people who use credit from banks reach to millions, trusting and customer-based lending process is very difficult. For this reason, techniques which are called credit scoring techniques and which measure whether given credit or not to customers using customer information are widely used among banks today. In this study, algorithms that can be used in credit scoring systems are examined. Within the scope of the study, a prototype implementation was developed that would facilitate the approval or denial of credit requests by customers. The developed application presents a comparison of commonly used machine learning methods with deep learning methods such as Classification Restricted Boltzmann Machine and Multilayer Artificial Neural Networks. Performance tests of the prototype application have been performed and positive results have been obtained showing the availability of the system.