Statistical machine learning methods applied to ATB data for credit risk modelling

Machine learning (ML) is a method of training a computer to learn from data and predict future outcomes based on existing patterns in the data. This project aims to utilize various ML methods as new and potentially better analytics and predictive tools in the area of credit risk management for ATB. Given that data quality and flows change over time, a new framework built on Google Cloud Platform to update the machine learning models will also be developed. Additionally, considering the possibilities that the ML models may favour certain subgroup, e.g., defined by race and geography, new strategies to test and correct for model fairness will be established. In summary, we would improve default risk prediction accuracy with the help of leading edge ML methods and expand the field of credit risk management by providing model updates through Google Cloud Platform and establishing model fairness strategies.

Faculty Supervisor:

Bei Jiang

Student:

Lisa Shulman

Partner:

ATB Financial

Discipline:

Mathematics

Sector:

Finance, insurance and business

University:

University of Alberta