Horizon CDT Research Highlights

Research Highlights

Exploring responsible applications of deep learning for consumer credit risk assessment

  Edwina Borteley Abam (2019 cohort)


To explore how to apply deep neural networks for consumer credit risk assessment in a responsible manner.

Research Questions:

  1. Can we benefit from using deep learning as a traditional tool for credit risk assessment?

  2. How can we design an interpretable deep neural network algorithm for responsible decision making in the lending sector?

Overview and Context:

With an increased demand for credit cards and other loans, lenders such as banks are faced with the challenge of effectively measuring the risk of lending to borrowers. The financial industry is continuously researching efforts to make 

Constructing machine learning credit scoring models to assess consumer credit risk automatically is a typical approach banks use to tackle the issue. In recent years, there has been a widespread adoption of deep neural network models in many application areas as object recognition, natural language processing and pose estimation. However, there has been minimal attention given to deep neural networks in the lending sector. The reason being their use in  decision making systems such as credit scoring raises questions of trust, accountabilty and fairness. The focus of this research is to explore how deep neural network models can be used to make lending decisions in a transparent way that promotes trust and ensures responsible use of the algorithm in the credit industry.