Application of Transformer Models to Raw Credit Bureau Files for Improved Credit Risk Modelling Performance

This research project will look at new ways to better understand and evaluate a person’s creditworthiness using advanced computer techniques. Traditionally, when banks or lenders decide if someone can be trusted with a loan, they look at specific numbers and data. However, there’s a lot of information in written reports that these traditional methods might miss. Our project will use a technology called Natural Language Processing (NLP) to read and analyze these reports in depth. By doing so, we hope to make credit evaluations more accurate. For the partnering organization, this means they can make better lending decisions, which can lead to increased profits and fewer bad loans.

Faculty Supervisor:

Vahab Khoshdel

Student:

Partner:

Wealthsimple Technologies

Discipline:

Computer science

Sector:

Information and cultural industries; Mining

University:

University of Manitoba

Program: