Dataset Quality Issues CT-AI Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   What is the primary purpose of ongoing data quality monitoring in an ML pipeline?

   • ATo detect and address data quality issues that may arise after the model is deployed.
   • BTo permanently fix all data quality issues automatically.
   • CTo replace the need for initial data cleaning before training.
   • DTo improve the model's algorithm without retraining.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   Which technique is used to automatically examine a dataset to summarize its structure, such as counting missing values, unique values, and data types for each column?

   • AData imputation
   • BData normalization
   • CData profiling
   • DData deduplication
   Reveal Solution Discussion Select an answer first
4. 13expert · hard

   A data scientist is working on a binary classification problem. The dataset has a 'label' column that was manually annotated by multiple annotators. The scientist discovers that for 5% of the records, the annotators disagreed on the label. The scientist wants to ensure the model is trained on reliable labels. Which approach is most appropriate?

   • AUse the majority vote label for all records
   • BReview the disagreed records and resolve them through a consensus process
   • CRemove all records where annotators disagreed
   • DIgnore the disagreements and use the first annotator's label
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   A data scientist is preparing a dataset for a model that predicts whether a customer will default on a loan. The dataset has a 'credit_score' column with 10% missing values. The scientist is considering two strategies: imputing with the mean or deleting the rows with missing values. The dataset is relatively small, and the missingness is not completely random. Which strategy is more likely to introduce bias?

   • ADeleting the rows is more likely to introduce bias because the missingness is not random
   • BImputing with the mean is more likely to introduce bias because it reduces variance
   • CBoth strategies are equally likely to introduce bias
   • DNeither strategy introduces bias
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   In a classification dataset, some instances are assigned the wrong class label. Which common data quality issue does this describe?

   • ADuplicate
   • BMissing value
   • CNoise
   • DLabel error
   Reveal Solution Discussion Select an answer first