Abstract

In this paper we present a novel approach to credit scoring of retailcustomers in the banking industry based on deep learning methods. We used RNNson fine grained transnational data to compute credit scores for the loanapplicants. We demonstrate that our approach significantly outperforms thebaselines based on the customer data of a large European bank. We alsoconducted a pilot study on loan applicants of the bank, and the study producedsignificant financial gains for the organization. In addition, our method hasseveral other advantages described in the paper that are very significant forthe bank.

 

Quick Read (beta)