Pragmatic segmentation-based credit risk management using Machine Learning

Historical financial crises like the 2007–2008 global financial crisis are a great example of the need for commercial institutions to predict the credit risk of their clients. Credit cards are a convenient and quick way to pay for products and services, but as debt increases over time, moreso with the pandemic and growing unemployment, the possibility of late payments increases. As the number of credit card users has been increasing, banks have been facing an escalating credit card default rate. This paper proposes a novel practical technique to segment customers by their predicted probability of defaulting on payments to help financial institutions assess risk before issuing credit cards. We have used the Taiwan credit card default dataset to present our approach and results. However, the proposed approach has been generalized and can be applied to credit card default datasets from other countries as well.