The Challenge.
- Many financial institutions have rules-based algorithms to determine whether a customer is eligible for a loan, often using payment history as the main factor.
- However, there are many other factors such as demographics that can impact the probability that a customer will default on a loan.
- Advanced Machine Learning algorithms like deep learning can be powerful at predicting default on loans, however the decisions are difficult to interpret and explain.
The Solution.
- Designed and built a Machine Learning model to predict the risk that a customer will default on a loan, using interpretable ML algorithms.
- Integrated the model into enterprise systems for automated mortgage approval within minutes using a few key pieces of information.
- Built reports to explain how the algorithm works and that can trace why a specific decision (approval or non-approval) was taken to ensure accountability.
The Results.
- Reduced the rate of mortgage default.
- Reduced the cost of assessing credit risk.
- Improved customer experience by enabling fast credit approval from anywhere.