AI0-001 Implementing AI Solutions 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.

• ✗

  Image classification model (e.g., CNN)

  Why it's wrong here

  A CNN classifies pixel patterns into fixed visual categories, so it cannot consume tabular purchase history or browsing sequences to rank products. It is tempting because recommendation output looks like labelling, and CNNs excel at image tasks; it would be correct if the platform recommended visually similar products from catalogue photographs.

• ✗

  Linear regression

  Why it's wrong here

  Linear regression predicts a continuous numeric value from input features, so it cannot rank or score discrete catalogue items from purchase and browsing signals. It is tempting because it is the standard first model for forecasting tasks such as predicting spend; it would be the right choice if the platform needed to estimate a customer's next order value.

• ✗

  Random forest classifier

  Why it's wrong here

  A random forest classifier predicts discrete labels, not ranked product preferences from purchase and browsing sequences; recommendation needs collaborative or content-based ranking. Random forests suit tabular classification such as churn or fraud. They are tempting because they handle mixed features well, but they output classes rather than relevance scores.

• ✓

  Collaborative filtering model (e.g., matrix factorization)

  Why this is correct

  Collaborative filtering exploits the interaction matrix between users and items, learning latent factors from purchase and browsing history to predict unseen preferences. This directly matches the scenario's reliance on behavioural signals rather than item content, making matrix factorisation the best-suited approach for personalised product suggestions.