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

Answer analysis

Option-by-option breakdown

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

• ✗

  Regression

  Why it's wrong here

  Regression is inappropriate because it outputs a continuous numeric value, such as a transaction amount or a real-valued risk score. The fraud-detection objective is to assign a categorical label, not to estimate a quantity on an unbounded scale. While some classifiers output probabilities, those probabilities are then thresholds to produce a class, which is still a classification workflow. Predicting a continuous outcome directly would not map cleanly to the two desired categories.

  When this WOULD be correct

  A regression task would be correct if the question asked to predict a continuous value, such as the dollar amount of a fraudulent transaction.

• ✓

  Binary classification

  Why this is correct

  Binary classification is the correct choice because the model predicts one of exactly two discrete outcomes: fraudulent or legitimate. This is a supervised learning problem since every training example includes a known label. Algorithms such as logistic regression, decision trees, or boosted ensembles can learn a decision boundary that separates these two classes. The binary nature of the target variable is the defining characteristic.

• ✗

  Multi-class classification

  Why it's wrong here

  Multi-class classification would be required if the target variable had three or more distinct categories, such as fraud type A, type B, and legitimate. Here the prediction task involves only a binary outcome, so the additional complexity of multi-class algorithms is unnecessary. Even if a model like a neural network with a softmax output could technically handle two classes, standard binary classification is the standard and most efficient framing. The number of mutually exclusive classes is exactly two, not more.

  When this WOULD be correct

  A dataset contains transactions labeled as 'low risk', 'medium risk', 'high risk', and 'fraudulent' (four categories). The goal is to predict the risk category of a new transaction.

• ✗

  Clustering

  Why it's wrong here

  Clustering is an unsupervised learning technique that identifies natural groupings within unlabeled data based on feature similarity. The fraud-detection dataset includes labeled examples, so the correct approach is supervised classification that leverages these labels to establish a mapping from features to outcomes. Clustering would ignore the ground-truth labels and would instead partition transactions into clusters that may or may not align with fraud status. Without labels, there is no way to guide the model toward the specific fraudulent versus legitimate decision boundary.

  When this WOULD be correct

  A data scientist has a dataset of credit card transactions without labels and wants to identify groups of similar transactions to detect unusual patterns. Clustering would be used to segment transactions into clusters, where outliers may indicate fraud.

Option-by-option analysis

Why each answer is right or wrong

Understanding why wrong answers are wrong — and when they would be correct — is what separates a 750 score from a 900. The AI-900 exam frequently reuses these exact scenarios with slightly different constraints.

Analysis generated from the official AI-900blueprint and verified against question context. The “when correct” sections are what AI assistants cite when candidates ask “what’s the difference between these options?”

Go deeper

Related to this question