Training and Evaluation

1. What is the primary purpose of a test set in the split-train-evaluate discipline?

To train the model further To validate model choices To estimate real-world performance To tune hyperparameters

2. Which metric would you prioritize when designing a fraud detection system where missing a fraudulent claim is costly?

Precision Accuracy Recall F1 Score

3. If a model shows high training scores but significantly lower test scores, what issue is it likely suffering from?

Underfitting Overfitting Low precision High variance

4. Why is accuracy not a reliable metric in the case of class imbalance, such as in a fraud detection scenario?

It does not account for true positives It can be misleading with high true negatives It fails to differentiate between fraud and legitimate claims It ignores false positives

Continue: When Not to Use an LLM → Review this lesson Retake quiz