Different AI Models and Development Processes AIE Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A fraud detection team has built a model to identify fraudulent transactions. The dataset is highly imbalanced: only 1% of transactions are fraudulent. The team evaluates the model using accuracy and sees 99% accuracy. What is the most important reason to use additional evaluation metrics?

   • AAccuracy does not reflect performance on the minority class
   • BAccuracy is too slow to compute on large datasets
   • CAccuracy is not a standard metric in the industry
   • DAccuracy cannot be used for classification models
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A bank wants to build a model to approve or reject loan applications based on applicant data. The bank has a large historical dataset with records of approved and rejected loans. The model must output a clear decision for each applicant. Which model type and selection criterion best fits this requirement?

   • ASupervised regression because it can output a probability score
   • BReinforcement learning because it can learn from approval outcomes over time
   • CSupervised classification because the dataset contains labeled outcomes
   • DUnsupervised clustering because it can find hidden applicant groups
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A team has built a model to predict whether a patient has a disease. The model has high precision but low recall. The medical team wants to ensure that no patient with the disease is missed. What should the team do?

   • AKeep the current threshold because precision is more important
   • BRaise the decision threshold to increase precision
   • CLower the decision threshold to increase recall
   • DSwitch to a different model type without tuning
   Reveal Solution Discussion Select an answer first
5. 29application · medium

   A startup wants to build a model to estimate the price of used cars based on features like mileage, age, and brand. They have a dataset with 10,000 labeled examples. Which model type is most appropriate?

   • AUnsupervised clustering model
   • BSupervised classification model
   • CSupervised regression model
   • DReinforcement learning model
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A city government wants to build a model to predict traffic congestion. They have collected years of traffic sensor data. The team is at the stage of defining the problem. What should they do first in this stage?

   • ADefine clear success criteria and the specific prediction target
   • BCollect additional data from external sources
   • CTrain a preliminary model to see what is possible
   • DDeploy a prototype to gather user feedback
   Reveal Solution Discussion Select an answer first