Objective 2.1 Recognize Relative Impact of Data Quality and Size to Algorithms AIP-210 Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   Which technique is commonly used to identify missing values and outliers in a dataset?

   • AModel deployment
   • BData profiling
   • CFeature scaling
   • DHyperparameter tuning
   Reveal Solution Discussion Select an answer first
2. 17expert · hard

   A data team is building a model to forecast demand for retail products. They have historical sales data, but the data for the last three months is incomplete because a new POS system was rolled out and some transactions were not recorded. They also notice that the 'product_category' field has several different spellings for the same category (e.g., 'Electronics', 'electronics', 'ELECTRONICS'). Which data quality issues are present, and which should they address first?

   • ACompleteness and accuracy; address accuracy first because incorrect values are more harmful
   • BConsistency and uniqueness; address consistency first because it is easier to fix
   • CCompleteness and consistency; address completeness first because missing recent data can bias the forecast
   • DAccuracy and timeliness; address timeliness first because the data is outdated
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   A dataset contains customer records where the same customer's age is listed as '35' in one system and '35 years' in another. Which data quality dimension is violated?

   • ATimeliness
   • BConsistency
   • CAccuracy
   • DCompleteness
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   How can missing values in a training dataset degrade a machine learning model's performance?

   • AThey can cause the model to learn patterns that do not generalize to new data.
   • BThey have no effect on model performance if the algorithm is linear.
   • CThey only affect the model's training time, not its accuracy.
   • DThey always increase the model's accuracy because less data means less noise.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A team is training a language model. They have 1 million sentences for training and observe that the validation loss is decreasing steadily. They have the option to add 500,000 more sentences, but they are concerned about the cost of annotation. Based on the learning curve, what should they consider?

   • AThe additional data will definitely improve the model because more data is always better
   • BThe additional data will have no effect because the model has already converged
   • CThe additional data may yield diminishing returns, so they should evaluate the cost-benefit trade-off
   • DThe additional data will cause overfitting because the model is already large
   Reveal Solution Discussion Select an answer first