Machine Learning Modelling for APP Scam Prevention

This document is a presentation that discusses the application of machine learning (ML) modeling in the prevention of Authorised Push Payment (APP) scams. It outlines the limitations of conventional rule-driven criteria typically used in the industry, which, while providing a robust baseline defense, may struggle against more complex fraudulent behaviors. The presentation details how an ML ensemble model can utilize advanced statistical algorithms to identify complex patterns, enhancing scalability, efficiency, and autonomy in fraud prevention. A significant challenge addressed in this project was the integration of the ML model into the existing fraud prevention system, which was achieved by exporting the models into the Predictive Model Markup Language format. The document also includes a case study demonstrating the successful implementation of this advanced analytical approach at Nationwide, illustrating the benefits of combining ML models with traditional methods for a multi-layered defense against APP scams.