Different AI Models and Development Processes AIE Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A credit card company is developing a fraud detection model. The cost of a false positive (blocking a legitimate transaction) is high, but the cost of a false negative (allowing fraud) is even higher. The model currently has high precision but low recall. The team must decide how to proceed. What is the best action?

   • ADeploy the model as is because high precision minimizes customer inconvenience
   • BUse an unsupervised anomaly detection model instead
   • CLower the classification threshold to increase recall, accepting more false positives
   • DRaise the classification threshold to increase precision further
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A manufacturing company wants to detect defective products on an assembly line using images. They have a large dataset of images labeled as 'defective' or 'non-defective'. The system must flag defects in real time. Which model type is most suitable?

   • AReinforcement learning model
   • BSupervised image classification model
   • CSemi-supervised model using only unlabeled images
   • DUnsupervised anomaly detection model
   Reveal Solution Discussion Select an answer first
4. 3expert · hard

   A retail chain wants to forecast daily sales for each store. They have historical sales data, but the data contains missing values and outliers due to holidays and system errors. The team is at the data preparation stage. What is the most important consideration before model training?

   • ADeploy the model immediately and rely on real-time corrections
   • BSelect a model that does not require data cleaning
   • CImpute missing values and handle outliers appropriately to avoid biased predictions
   • DUse only the most recent data to avoid missing values
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A team has developed a spam email classifier. During evaluation, they notice that the model correctly identifies most spam emails but also marks many legitimate emails as spam. Which evaluation metric should the team focus on to reduce this problem?

   • AAccuracy
   • BRecall
   • CPrecision
   • DF1 score
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A financial firm wants to build a model to detect unusual trading activity. The firm has a large dataset of normal transactions but very few examples of fraudulent activity. The cost of missing fraud is high, but the firm also wants to minimize false alarms. Which model selection and evaluation approach is best?

   • AUse supervised classification with the available fraud labels and evaluate with accuracy
   • BUse reinforcement learning to learn a trading policy
   • CUse unsupervised anomaly detection to identify deviations from normal behavior and evaluate with precision and recall
   • DUse supervised regression to predict a fraud score
   Reveal Solution Discussion Select an answer first