Different AI Models and Development Processes AIE Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A startup wants to build a recommendation system for a new video streaming service. They have no user interaction data yet, but they have a catalog of movies with genres, actors, and directors. They need to provide initial recommendations. Which approach is most appropriate?

   • ACollaborative filtering using user ratings
   • BReinforcement learning to learn user preferences over time
   • CContent-based filtering using item features
   • DSupervised regression on user ratings
   Reveal Solution Discussion Select an answer first
2. 17expert · hard

   A team is building a model to predict customer churn. They have a highly imbalanced dataset where only 2% of customers churn. The business wants to identify as many churners as possible, but the model must not overwhelm the retention team with false positives. Which evaluation strategy best balances these conflicting requirements?

   • AUse the ROC-AUC score as the sole metric
   • BOptimize for accuracy on the overall dataset
   • COptimize for recall only to catch all churners
   • DUse the F1 score and consider precision-recall trade-offs
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A team has trained a model to detect fraudulent transactions. The model achieves high accuracy on the training data but performs poorly on a held-out test set. What is the most appropriate next step in the development process?

   • ADeploy the model because training accuracy is high
   • BCollect more data only if the test set is too small
   • CSwitch to an unsupervised model to avoid the problem
   • DIterate on the model by adjusting hyperparameters or using regularization to reduce overfitting
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A city government wants to use AI to predict traffic congestion. They have years of traffic sensor data but no labels indicating congestion levels. The team needs to identify patterns of congestion. Which development process stage should they focus on first?

   • AData collection and labeling to create a supervised dataset
   • BData exploration and feature engineering to prepare for unsupervised learning
   • CModel training with a supervised classifier
   • DDeployment of a pre-trained model
   Reveal Solution Discussion Select an answer first
6. 20expert · hard

   A hospital wants to predict patient readmission risk within 30 days. They have a small dataset of 500 patients with many clinical features. The model must be interpretable for clinicians and achieve high recall to avoid missing at-risk patients. Which approach best balances these constraints?

   • ALogistic regression with threshold tuning to favor recall
   • BRandom forest with default threshold
   • CDeep neural network with dropout regularization
   • DSupport vector machine with RBF kernel
   Reveal Solution Discussion Select an answer first