Verify Data Quality for Go/no-Go Decision to Conduct Data Preparation PMI-CPMAI Practice Questions (Page 1)

Questions 1–5

1. 1application · medium

   A data engineer applied min-max scaling to numerical features and one-hot encoding to categorical features in a customer-churn dataset. Before approving the dataset for model training, the data scientist reviews the transformed data and notices that the 'income' column has values ranging from -0.5 to 1.5, while all other scaled features are between 0 and 1. What should the data scientist do first?

   • ARe-run the scaling step and verify the output range for every feature.
   • BDocument the anomaly and proceed, noting it in the data quality report.
   • CRemove the 'income' feature from the dataset to avoid the scaling issue.
   • DProceed with training because scaling does not need to be uniform across all features.
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   What does it mean for a dataset to be 'representative' of the target population?

   • AThe dataset's characteristics mirror the distribution of the target population
   • BThe dataset is free from any missing values
   • CThe dataset is collected from a single source
   • DThe dataset contains every possible record from the population
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   In a final data quality assessment, which activity is primarily used to evaluate the 'accuracy' dimension?

   • AChecking for missing values in critical fields
   • BConfirming the data was collected within the last month
   • CComparing sample records against source-of-truth data
   • DVerifying that date formats are uniform across the dataset
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A financial services company is preparing a loan-approval model. The final data quality assessment shows that the 'credit_score' column has 10% missing values, and the 'annual_income' column has values that are inconsistent with the 'employment_status' column (e.g., unemployed applicants with very high income). What should the team do before the go/no-go decision?

   • ADrop the credit_score column because it has too many missing values.
   • BImpute the missing credit scores and correct the income inconsistencies before proceeding.
   • CDocument the issues and proceed, noting them in the model documentation.
   • DProceed to training because the model can handle missing values and inconsistencies.
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A telecom company is building a model to predict customer churn. The training data includes customers from the past 24 months, but the company recently launched a new pricing plan that has attracted a different customer demographic. The model will be used to predict churn for the next 12 months. The team has a tight deadline and cannot collect new data. What is the most appropriate action?

   • AUse the historical data but incorporate the new pricing plan as a feature to capture the change.
   • BDelay the go decision and explain to stakeholders that the data is not representative of the new customer base.
   • CTrain the model on the historical data and restrict its use to customers on the old pricing plans.
   • DProceed with training because the data is the most recent available and the model can be retrained later.
   Reveal Solution Discussion Select an answer first