Dataset Quality Issues CT-AI Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A data engineering team is building a data quality monitoring system for a real-time recommendation engine. The data arrives in streaming batches, and the team wants to detect sudden changes in data distribution that could indicate a data quality issue. Which technique is most effective for this purpose?

   • AManually inspect a random sample of each batch
   • BUse a static threshold for all data quality metrics
   • CPerform a full data quality audit once a month
   • DCompute summary statistics on each batch and compare them to a rolling baseline
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A team is building a model to predict employee attrition. The dataset contains a 'satisfaction_score' column where 30% of the values are missing. The team decides to impute the missing values with the mean satisfaction score. What is the most likely consequence of this approach?

   • AThe model will be more accurate because the mean is a good representative value.
   • BThe model will overfit because the imputed values are too similar.
   • CThe model will underestimate the variability of satisfaction scores, potentially reducing its ability to detect patterns.
   • DThe model will be unbiased because the mean is an unbiased estimator.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   A dataset contains records with conflicting values for the same attribute across different sources (e.g., one source lists a product's price as $10 and another as $12). Which data quality dimension is most directly compromised?

   • ARelevance
   • BAccuracy
   • CCompleteness
   • DConsistency
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   How can poor data quality, such as systematic errors in the training data, most directly affect an ML model?

   • AIt can lead to a model that learns incorrect patterns and makes biased or unreliable predictions.
   • BIt has no effect as long as the model algorithm is well-chosen.
   • CIt only affects the model's ability to handle new data, not its performance on training data.
   • DIt only affects the model's training speed, not its final performance.
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   Which of the following is an approach used for ongoing monitoring of data quality in an ML pipeline?

   • AIgnoring data quality after the model is deployed.
   • BSetting up automated alerts when data quality metrics exceed predefined thresholds.
   • CManually reviewing the entire dataset every day.
   • DPerforming a one-time data quality assessment before model training.
   Reveal Solution Discussion Select an answer first