Design Use Cases CERTIFIED-DATA-SCIENCE-PRACTITIONER Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   A team is establishing success criteria for a model that predicts equipment failure. Which criterion is most appropriate?

   • AThe model must use at least 10 features.
   • BThe model must be approved by the project manager.
   • CThe model must be better than random guessing.
   • DThe model must achieve an F1 score of at least 0.85 on the validation set.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   A team is defining the goals for testing a new fraud detection model. Which statement best articulates a clear testing objective?

   • AThe model should be fast enough for real-time use.
   • BThe model should not make too many mistakes.
   • CThe model should achieve a precision of at least 0.90 on the holdout test set.
   • DThe model should be better than the previous version in some way.
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A telecom company is deploying a model to predict customer churn. The business wants to maximize the number of customers retained while minimizing the cost of retention offers. The model's output is a probability of churn. Which success criterion best balances these competing goals?

   • AMaximize recall only
   • BMaximize overall accuracy
   • CMaximize precision only
   • DMaximize the area under the precision-recall curve (PR-AUC)
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A retail company has a small dataset of 5,000 customer transactions and wants to test a churn prediction model. They need a testing method that uses all data for both training and validation while providing a robust estimate of performance. Which method should they choose?

   • AK-fold cross-validation
   • BA/B testing on live traffic
   • CBootstrap sampling
   • DHoldout validation with a 70/30 split
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A fraud detection model is being tested. The team wants to ensure it handles typical transactions, unusual but legitimate transactions, and deliberate attempts to evade detection. Which set of test cases best covers these conditions?

   • AOnly transactions from the last month
   • BTransactions with only the most common features
   • CTransactions that are all labeled as fraud
   • DNormal transactions, transactions with missing fields, and transactions with extreme amounts
   Reveal Solution Discussion Select an answer first