Data Analytics for Credit Risk Models in Retail Banking: a new era for the banking system

Abstract

Given the nature of the lending industry and its importance for global economic stability, financial institutions have always been keen on estimating the risk profile of their clients. For this reason, in the last few years several sophisticated techniques for modelling credit risk have been developed and implemented. After the financial crisis of 2007-2008, credit risk management has been further expanded and has acquired significant regulatory importance. Specifically, Basel II and III Accords have strengthened the conditions that banks must fulfil to develop their own internal models for estimating the regulatory capital and expected losses. After motivating the importance of credit risk modelling in the banking sector, in this contribution we perform a review of the traditional statistical methods used for credit risk management. Then we focus on more recent techniques based on Machine Learning techniques, and we critically compare tradition and innovation in credit risk modelling. Finally, we present a case study addressing the main steps to practically develop and validate a Probability of Default model for risk prediction via Machine Learning Techniques