How would you design a model to detect upcoding or fraudulent billing patterns in a highly imbalanced dataset of medical claims at Machinify?
