A data scientist trains a classification model to predict whether an email is 'phishing' or 'legitimate'. The model achieves 99% accuracy on the training data but only 68% accuracy on the test data. Which action is most likely to help improve the model's generalization performance?

Increase the number of training epochs significantly.

More training epochs typically increase the risk of overfitting, as the model can memorize the training data more thoroughly. This would likely widen the accuracy gap.

Remove some of the training data to make the dataset smaller.

Reducing the amount of training data usually exacerbates overfitting because the model has fewer examples to learn from, making it more likely to memorize the remaining data.

Add more layers and neurons to the neural network.

Increasing model complexity by adding more parameters generally increases the risk of overfitting, especially when the dataset is relatively small.

What does this AI-900 question test?

Static NAT maps one inside address to one outside address.

What is the correct answer to this question?

The correct answer is: Apply regularization techniques such as L1 or L2 regularization. — The large gap between training and test accuracy indicates overfitting, where the model has memorized the training data rather than learning general patterns. Applying regularization (such as L1/L2 penalties) constrains the model to reduce its complexity, thereby improving its ability to generalize to unseen data. Increasing complexity or training time would worsen overfitting, and reducing data would likely make it worse.

What should I do if I get this AI-900 question wrong?

Then try more questions from the same exam bank and focus on understanding why the wrong options are tempting.