Using Machine Learning to Detect Financial Fraud, 2021

This presentation was part of the Brown Bag livestreamed lecture series on masterclass.sg.

Financial fraud occurs rarely but costs economies billions of dollars. Consequently, fraud detection is challenging but plays an important role in safeguarding society. This talk will break down how machine learning helps make fraud detection easier through unlocking data for flagging fraud and providing algorithms focused on rare event detection.

The lecture covers:

1. What financial fraud entails.
2. Why it is difficult to detect, as well as some ways that have been used for detection previously.
3. A modern approach to detecting fraud using machine learning to measure the content of annual reports.
4. A modern approach using machine learning to replace the traditional regression struture used for fraud detection.

Slides are available in an interactive web-based format (reveal.js) and in PDF. A video of the talk is available publicly, hosted by HeadHunt.

More details are available in my paper with W. Brooke Elliott and Nerissa Brown, published in the Journal of Accounting Research. A copy is available publicly on SSRN: