Historically banks have granted loans to applicants based on their income and credit scores which limits the growth potential of their lending business. For example, this framework will not necessarily accommodate for good income earners with no credit history. Moreover, the framework is not perfect since a percentage of the approved applicants will become delinquent.
Luckily, banks and lenders have historical records of the customers who were granted loans and whether they defaulted or not. These records will include many customer attributes including their credit scores, the purpose of the loans, interest rates, their age, etc. By training a Machine Learning model, using the customer attributes as inputs and the loan status (defaulted or not) as the target to predict, lenders can understand what actually caused their customers to default.
The Machine Learning model can then be applied to the applicants who were originally rejected in order to predict whether they would default or not. It can also be applied to new applicants who might have passed the original screening.
This approach not only allows lenders to grow their business, but it will also improve their profitability by reducing the percentage of delinquent loans. A better understanding of the drivers of defaults can also help banks educate customers so that they can improve their chances of being approved as well as honoring their repayments.