PA-VALIDATION

What is the main purpose of splitting a dataset into training and test sets?

• To reduce the data size

• To improve code readability

• To evaluate how the model performs on unseen data

• To increase training time

What is the role of the validation set in machine learning?

• Final model testing

• Hyperparameter tuning

• Saving the model

• Increasing training data

What happens if the model is evaluated on the training data only?

• Results are always accurate

• Model will generalize well

• It speeds up the model

• Model may overfit and results may be misleading

Which of the following is NOT a correct split ratio?

• 80% train, 20% test

• 70% train, 30% test

• 100% train, 0% test

• 60% train, 20% validation, 20% test

What does accuracy measure?

• Ratio of correct predictions

• Ratio of false negatives

• Speed of model

• Model size

In a confusion matrix, True Positives (TP) are:

• Negative examples incorrectly predicted as positive

• Positive examples correctly predicted as positive

• Negative examples correctly predicted

• Unused data

Precision is calculated as:

• TP / (TP + FP)

• TP / (TP + FN)

• TN / (TN + FN)

• FP / (FP + TP)

F1-score is useful when:

• Data is balanced

• You want to ignore false negatives

• You want a balance between precision and recall

• Model runs fast

The confusion matrix helps you:

• Tune code

• Speed up runtime

• Understand detailed prediction performance

• Store the model

In a spam filter, false positives are:

• Spam detected as spam

• Good emails marked as spam

• Spam ignored as good

• No emails