Fraud Definitions and Rigorous Labeling are Vital to the Public Sector

Record-high unemployment claims during the COVID-19 pandemic have increased the risk of unemployment insurance fraud, costing the public sector hundreds of millions of dollars and delaying delivery of benefits to people in need. Appropriately defining fraud and labeling data underpins machine learning models that effectively identify what is fraud and what is a legitimate claimant.

Staying ahead of accelerating identity fraud requires public sector administrators to make informed decisions about where and how to invest in protections. This white paper provides a framework for learning about the critical nature of fraud definitions and labeling that will lead to better stakeholder outcomes:

• Why fraud definitions are important in the public sector
• How fraud labels are used in models
• Why accurate labels are important
• The pros and cons of techniques that are used to label fraud
• What the best practices are for defining and labeling fraud

Abraham Lincoln said, “Give me six hours to chop down a tree and I will spend the first four sharpening the ax.” Defining and labeling fraud is the equivalent of sharpening the ax. When done right, the work of stopping fraud will, over time, become more precise, take less time, and be more productive for everyone other than the bad actors.