Knowledge Quiz

1.According to the article, what is the primary reason why fraud detection is often misunderstood as a machine learning problem?

2.Why might a fraud detection model with 'excellent' offline evaluation still lead to a 'terrible operational outcome' in production?

3.What does the article suggest is the 'real question' for a fraud detection model, rather than 'can the model separate classes in a notebook?'

4.Which of the following is NOT listed as a reason why good offline metrics can mislead in fraud detection?