There are a variety of insurance frauds, for example:

  • Fake claims
  • Fake policies
  • Identity theft

How to detect insurance frauds

The idea is to identify suspicious claims before they are damage, theft or loss is paid.

If we have labels, we can fit a classification model to classify wether a claim is a fraud or not. If we don’t have labels, which is pretty common in the fraud detection case, then we can use an unsupervised method such as contrastive learning.