Test Creation CERTIFIED-DATA-SCIENCE-PRACTITIONER Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   Why is it important for test data to be independent of the training data?

   • ATo ensure the model has enough data to learn from
   • BTo make the test set more representative of the training set
   • CTo reduce the size of the dataset
   • DTo avoid overfitting and get an unbiased estimate of performance
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   A team has completed testing a model and is documenting the results. They found that the model performs well on average but fails on a specific demographic subgroup. They are unsure whether to include this finding in the report. What is the best course of action?

   • ARetest the model on a different subgroup to see if the issue persists.
   • BInclude the finding, as it is important for transparency and further analysis.
   • CExclude the finding because it might alarm stakeholders.
   • DInclude only the overall metrics to keep the report concise.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   Which component is typically part of a test harness?

   • AA mechanism to load test data and run the model
   • BA deployment pipeline
   • CA hyperparameter optimizer
   • DA tool for data augmentation
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   What is a key requirement for test data to be considered appropriate for model validation?

   • AIt must be larger than the training set
   • BIt must contain only edge cases
   • CIt must be collected after the model is deployed
   • DIt must be representative of the real-world data the model will encounter
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A team is testing a fraud detection model. They have a test set with 10,000 transactions, but only 50 are fraudulent. To ensure the model is robust, they decide to create test cases that include synthetic fraudulent transactions that mimic known fraud patterns but with slight variations. Which test design principle are they applying?

   • AEnsuring the test set is independent from the training data.
   • BIncluding edge cases and potential failure modes that might not be well-represented in the original data.
   • CReducing the test set size to speed up execution.
   • DEnsuring the test set is representative of the real-world class distribution.
   Reveal Solution Discussion Select an answer first