Making Predictions Under Lending Regulat...

One particular difficulty when working in lending is being subject to a variety of different regulatory requirements. This has many implications when working with data - the requirements often affect the types of models you can build or restrict which predictors you can include when modelling. In this talk we present how the Fair Credit Reporting Act (FCRA) impacted our loan application accept / reject modelling. In particular, the FCRA requires explicit reasons for a declined application, which we satisfied through a thoughtful use of ensemble methods.