Artificial Intelligence and Machine Learning in Credit Risk Analytics: Present, Past and Future

Douglas Dwyer, Tony Hughes and Ashit Talukder

Algorithms have always played a key role in managing credit risk. Over time, structured data has become bountiful, computer hardware more powerful and software more user-friendly. As a result, the sophistication of algorithms has increased and will continue to increase. Unstructured data has become more accessible, and such information can be analysed using techniques taken from the artificial intelligence (AI) field. In this chapter, we review the state of play for AI and machine learning (ML) in credit risk management, how we arrived there and where we are going. We show that the biggest improvements in credit analytics are likely to arise from using AI and ML to bring new information into risk assessment. Furthermore, whether or not models are intuitive, transparent and reasonable will continue to be important in many (but not necessarily all) aspects of risk analytics.

Lending was active in ancient Greece, and during these ancient times collateral was used to manage credit risk. In the early 1800s, the Rothschilds “created” international banking by establishing an international network for transferring money. From 1816, the Philadelphia Saving Funds Society provided access