Courseiva

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

A data scientist trains a regression model to predict housing prices. The model uses polynomial features up to degree 5. It achieves an R-squared of 0.95 on the training set but only 0.60 on the test set. Which problem is the model most likely experiencing?

⚠ Common exam trap

A common mix-up: candidates confuse overfitting with underfitting, but the key indicator is the large gap between high training performance and low test performance, not uniformly low performance.

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

Overfitting

The model performs exceptionally well on the training data (R-squared 0.95) but poorly on the test data (R-squared 0.60), which is the classic symptom of overfitting. Using polynomial features up to degree 5 introduces high model complexity, causing the model to learn noise and specific patterns in the training set that do not generalize to unseen data.

Answer analysis

Option-by-option breakdown

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

  • Underfitting

    Why it's wrong here

    Underfitting reflects a model with excessive bias, such as a linear regression applied to inherently nonlinear data. It would produce high error on both the training and test sets, because the model fails to capture the underlying relationship even in the data it has seen. In contrast, the scenario describes strong training performance with weak test performance, which is the opposite of underfitting.

  • Overfitting

    Why this is correct

    Overfitting occurs when a complex model, such as a high-degree polynomial or a deep regression tree, memorizes the noise and idiosyncrasies of the training data. This yields near-perfect training predictions, but the model does not generalize to new data, producing a large train–test performance gap. The hallmark is high variance and low bias on training, with poor performance on validation or test sets.

  • Data leakage

    Why it's wrong here

    Data leakage arises when information from the test set inadvertently influences training, such as scaling features using statistics computed from the full dataset before a split. If leakage were present, we would typically see artificially high performance on the test set as well, because the model already 'knows' the answer. The observed low test performance therefore contradicts the typical leakage signature.

  • Multicollinearity

    Why it's wrong here

    Multicollinearity refers to high correlations among predictor variables, which inflates the variance of coefficient estimates and makes individual effects difficult to isolate. While it can make models unstable, it does not inherently create a systematic large gap between training and test performance unless combined with overfitting or extrapolation. The reported train–test discrepancy is best explained by the model fitting noise, not by collinearity among features.

About these practice questions

Courseiva writes every AI-900 question from scratch — 985 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →

How Courseiva writes practice questions · Editorial policy

JA

Written by Johnson Ajibi, MSc IT Security

Senior Network & Security Engineer · founder of Courseiva

This AI-900 practice question is part of Courseiva's free Microsoft certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI-900 exam.