Mislabeled Data in Datasets CT-AI Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A data science team is training a model to classify support tickets. They suspect that some tickets are mislabeled, but they are unsure of the extent. They have a limited budget and need to improve model performance. Which combination of detection and handling strategies is most cost-effective?

   • AManually review the entire dataset to ensure all labels are correct
   • BUse a robust learning technique and ignore the mislabels
   • CRemove all tickets that the model predicts with low confidence
   • DUse a model-based approach to flag suspicious tickets, then manually review only those tickets
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which of the following is a model-based approach to detect mislabeled data?

   • ATraining a model and examining instances where the model's prediction disagrees with the label
   • BManually reviewing a random sample of the dataset
   • CChecking for duplicate rows in the dataset
   • DComputing the mean and standard deviation of each feature
   Reveal Solution Discussion Select an answer first
4. 8application · medium · select all that apply

   A data science team is working on a dataset with suspected mislabeled instances. They want to use a combination of detection techniques to increase confidence. Which of the following are valid detection techniques for mislabeled data? (Select all that apply.)

   • AManual review of a random sample
   • BComparing labels to an external authoritative source
   • CRemoving all instances with missing values
   • DStatistical outlier analysis on feature values
   • EModel-based approach using prediction confidence
   Check answers Discussion Select an answer first
5. 9foundation · easy

   How can mislabeled data in the training set affect a supervised ML model?

   • AIt can degrade model accuracy by teaching the model incorrect patterns
   • BIt has no effect because the model learns from features only
   • CIt only affects the test set, not the training set
   • DIt always improves model generalization due to added noise
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A team is building a sentiment analysis model for product reviews. They find that a significant number of reviews are mislabeled due to a bug in the data collection script. The team has a limited budget and cannot afford to relabel all the data. Which handling strategy would best preserve model performance while staying within budget?

   • ARemove all mislabeled instances from the dataset
   • BIgnore the issue because the model will learn to ignore the noise
   • CUse a robust loss function that is less sensitive to label noise
   • DIncrease the model's complexity to better fit the noisy labels
   Reveal Solution Discussion Select an answer first