Data Quality and Its Effect on the ML Model CT-AI Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A data science team is preparing a dataset for a model that predicts employee attrition. They have two candidate datasets: Dataset A has 99% completeness and 98% accuracy, but was collected 3 years ago. Dataset B has 95% completeness and 90% accuracy, but was collected last month. The company's workforce policies have changed significantly in the past year. Which dataset is MORE suitable, and why?

   • ADataset B — the lower completeness is acceptable because the model can impute missing values.
   • BDataset B — timeliness is more critical because the workforce dynamics have changed.
   • CDataset A — the higher accuracy will compensate for the age of the data.
   • DDataset A — higher completeness and accuracy are the most important quality dimensions.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   Which data quality dimension measures whether data values fall within the expected range or format?

   • AAccuracy
   • BValidity
   • CTimeliness
   • DConsistency
   Reveal Solution Discussion Select an answer first
4. 18application · medium · select all that apply

   A data science team is evaluating the quality of a dataset for a model that predicts loan default. They have identified the following characteristics. Select ALL that indicate a data quality problem that could negatively impact the model.

   • AThe 'age' column contains values ranging from 18 to 100, which is a valid range.
   • BThe 'default' label is correct for 95% of records, but 5% are mislabeled.
   • C15% of the 'income' values are missing.
   • DThe data was collected from a single branch in one city.
   Check answers Discussion Select an answer first
5. 19foundation · easy

   What is a typical consequence of training an ML model on data with many duplicate records?

   • AThe model will be completely unaffected
   • BThe model will always underfit
   • CThe model will have higher accuracy on new data
   • DThe model may overfit to the duplicated examples
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A weather forecasting team is building a model to predict rainfall. The dataset contains a 'temperature' column where some values are recorded as -999, which is a placeholder for missing data. The team wants to handle these values. Which approach is MOST appropriate?

   • AReplace -999 with NaN and apply an imputation method that accounts for the temporal context.
   • BRemove all rows containing -999 from the dataset.
   • CTreat -999 as a valid temperature and include it in training.
   • DReplace -999 with the mean temperature of the dataset.
   Reveal Solution Discussion Select an answer first