Objective 2.1 Recognize Relative Impact of Data Quality and Size to Algorithms AIP-210 Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   Which of the following is an example of using visualization to assess data quality?

   • ATraining a model and evaluating its accuracy on a test set.
   • BPlotting a histogram of a feature to see its distribution and spot unusual spikes.
   • CSplitting the data into training and validation sets.
   • DCalculating the mean and standard deviation of a feature.
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A data scientist is working with a dataset where the 'age' feature has 15% missing values. They decide to impute the missing values with the median age. What is the most likely impact of this imputation on a linear regression model?

   • AIt will have no effect on the model because the missing values are random
   • BIt will increase the variance of the age feature, making the model more sensitive to outliers
   • CIt will preserve the central tendency of the feature but may reduce the model's ability to capture the true relationship for those records
   • DIt will completely eliminate any bias caused by the missing data
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A data scientist is building a churn prediction model from a customer database. During profiling, they find that the 'last_purchase_date' field is missing for 30% of records, and for another 20% the date is recorded as '0000-00-00'. The 'customer_region' field has inconsistent values like 'NA', 'N/A', 'North America', and 'north america'. Which data quality dimensions are most directly affected by these findings?

   • APrecision and recall
   • BAccuracy and timeliness
   • CCompleteness and consistency
   • DValidity and uniqueness
   Reveal Solution Discussion Select an answer first
5. 4expert · hard

   A team is building a sentiment analysis model for product reviews. They have two data sources: Source A has 2 million reviews scraped from the web, but many are duplicates and some are mislabeled due to automated sentiment detection. Source B has 20,000 reviews that were manually labeled by experts. The team has a limited budget and can only use one source for training. Which source should they choose, and why?

   • ASource A, because the sheer volume will help the model generalize better
   • BSource B, because smaller datasets always train faster and are easier to tune
   • CSource B, because the higher label accuracy will lead to a more reliable model despite the smaller size
   • DSource A, but only after removing duplicates and mislabeled reviews
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   What does a learning curve typically show about the relationship between training data size and model performance?

   • APerformance initially improves with more data, but eventually plateaus as additional data yields diminishing returns.
   • BPerformance improves linearly with every additional data point, without limit.
   • CPerformance is independent of data size once a model reaches a certain complexity.
   • DPerformance decreases as more data is added because the model becomes overfitted.
   Reveal Solution Discussion Select an answer first