AIF-C01 Fundamentals of AI and ML Practice Question

Answer analysis

Option-by-option breakdown

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

• ✗

  Use reinforcement learning to reward correct email topic assignments

  Why it's wrong here

  Reinforcement learning needs an environment, actions, and reward signals across sequential decisions. There is no interactive environment or reward function in the email corpus. Applying it would not naturally surface groupings and would require a defined objective that the scenario does not provide, making it an unsuitable choice.

• ✗

  Perform regression on email length to predict customer churn

  Why it's wrong here

  Regression predicts a continuous target, and here the target would be churn rather than email grouping. This approach also assumes a labeled churn outcome exists, which the scenario does not mention. It ignores the stated goal of discovering natural groupings and reducing dimensionality, so it does not address the requirement.

• ✓

  Apply unsupervised learning techniques such as clustering and dimensionality reduction

  Why this is correct

  With no labels, unsupervised learning is the appropriate paradigm. Clustering algorithms can group emails by textual similarity, and dimensionality reduction techniques such as principal component analysis or t-SNE can compress the feature space for visualization and downstream analysis. This directly satisfies the goal of discovering groupings without predefined categories.

• ✗

  Train a supervised classifier on the emails using sentiment labels

  Why it's wrong here

  Supervised classification requires labeled examples, and the scenario explicitly states the emails have no labels. Creating labels first would be a separate annotation project, not the described goal. The company wants to discover structure, so applying a supervised classifier is both infeasible without labels and misaligned with the objective.

Go deeper

Related to this question