Developing artificial intelligence models to reduce the risk of loan default for Great Southern Bank (formerly CUA)

Problem

Great Southern Bank (formerly CUA) is the largest member-owned financial institution in Australia, offering banking, health insurance and general insurance to more than 550,000 Australians. Great Southern Bank approached Max Kelsen to investigate the application of modern machine learning applied in assessing credit risk of loan holders.

Solution

Max Kelsen developed a selection of models to predict the risk of loan default up to 6 months before the default, using a variety of complex data sources, state of the art models, and the same explainability techniques used in our cancer and genomics research projects. These inputs included data from sources which had not been utilised before. Unlocking this data enabled Great Southern Bank's risk professionals to reflect, assess and test their own modelling against modern machine learning techniques.

Outcomes

Better understanding of the risk of the personal loans portfolio and propensity of loans to default may assist in reducing the risk of members defaulting on loans, allowing Great Southern Bank to engage their members more effectively and provide assistance earlier in the process for those dealing with financial hardship. By using AI and explainability techniques, Great Southern Bank also gains insight into the predictors of default and how modelling techniques can address bias.