Challenges and Opportunities of Using Fraud Labels

On one hand, the key objective for merchants is increasing customer acquisition, customer retention and growing revenues, and on the other hand, merchants need to accurately prevent and detect fraudsters. In the process of balancing fraud prevention and customer acquisition, fraud labels have been playing a pivotal role in helping merchants to accurately predict fraud and protect customers. However, challenges remain as to how we can use fraud labels accurately? How to reduce bias? As well as how to further enhance the labels for machine learning model training.

Learning Objectives:

• How do we define fraud?
• How do we know if the fraud label is reliable?
• How do we reduce label bias and noise?
• How do we find ‘proxy labels’ when no actual labels exist?

Some content is hidden, to be able to see it login here Login