How would you structure an analysis of a massive dataset of healthcare claims to find fraudulent billing?
