A retailer has historical transactions labeled as fraudulent or legitimate and wants to predict fraud for new transactions. Which learning approach matches this training data?

1. Object lifecycle management
2. Supervised learning
3. Random assignment
4. Unsupervised clustering only