Definition #
ML classifiers trained on verified fraud cases.
Key Characteristics #
- Transaction labeling
- Feature engineering
- Model retraining
- Explainability
Why It Matters #
Detects 98% of known fraud patterns (FICO).
Common Use Cases #
- Credit card fraud
- Insurance claims
- Healthcare billing
Examples #
- SAS Fraud
- Featurespace
- Feedzai
FAQs #
Q: Data requirements?
A: Minimum 10k labeled fraud samples.
Q: Novel fraud?
A: Combines with unsupervised methods.