Describe fundamental principles of machine learning on Azure →hardMultiple ChoiceObjective-mapped

AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

A real estate company trains a model to predict house prices. They evaluate it on a test set of 100 houses. The model predictions have a mean absolute error (MAE) of $5,000 and a root mean squared error (RMSE) of $20,000. What does the large difference between MAE and RMSE indicate about the model's errors?

⚠ Common exam trap

A common mix-up: candidates assume a large RMSE always means the model is poor overall, but the question tests the understanding that a large gap between RMSE and MAE specifically reveals the presence of outliers with large errors, not uniform inaccuracy.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

The model has many small errors and a few large errors.

The mean absolute error (MAE) of $5,000 and root mean squared error (RMSE) of $20,000 show a large discrepancy because RMSE squares errors before averaging, which heavily penalizes large deviations. Since RMSE is four times larger than MAE, this indicates that while most predictions are close (small errors), there are a few predictions with very large errors that inflate the RMSE. This pattern is classic for a model that performs well on most houses but fails badly on a few outliers.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✓
The model has many small errors and a few large errors.
Why this is correct
The mean absolute error (MAE) and root mean squared error (RMSE) differ because squaring errors amplifies large residuals disproportionately. A large gap between RMSE and MAE indicates that most predictions have small errors, but a relatively few predictions have very large errors, which dominate the squared-error term. This pattern often arises when a model fits typical properties well but fails on outliers, such as luxury homes or distressed sales.
✗
The model consistently overestimates prices.
Why it's wrong here
Both MAE and RMSE are non-negative error magnitudes that discard the direction of the prediction error. A consistent overestimation means the errors are systematically one-sided, but both metrics would increase in proportion because they ignore sign; the gap between them does not reveal bias. The gap specifically measures the spread of error magnitudes, not whether predictions are systematically too high or too low.
✗
The model has a high bias and low variance.
Why it's wrong here
High bias and low variance describe a model that underfits and makes similar, systematic errors across the whole dataset. In that case, errors are relatively uniform, so the squared errors are not disproportionately large for a few points, keeping MAE and RMSE close together. A large gap between MAE and RMSE actually suggests high variance and the influence of outliers, not a high-bias/low-variance structure.
✗
The model is perfectly accurate.
Why it's wrong here
If the model were perfectly accurate, every predicted price would exactly equal the actual price, making every individual error zero. Consequently, both the MAE and RMSE would be zero, and there would be no gap between them to analyze. The presence of a non-zero gap between MAE and RMSE implies non-zero errors, so perfect accuracy is impossible.

Go deeper

AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

Why each option matters

Option-by-option breakdown

Related to this question