Model Performance Monitoring Quiz

What you'll be asked

1. Which of the following metrics is most commonly used to detect 'Data Drift' in numerical input features?
2. What is the primary purpose of implementing model monitoring in a production ML pipeline?
3. If your model's prediction accuracy drops, but the input data distribution remains identical to the training set, what is the most likely cause?
4. You notice that a feature's values in production have shifted significantly compared to the training set, but the model's overall accuracy remains stable. What is the best course of action?
5. When monitoring a model that predicts a continuous target variable, why is 'Mean Absolute Error' (MAE) monitoring preferred over 'Mean Squared Error' (MSE) in systems with outliers?

Answers and explanations come after you submit.