Verify Data Quality for Go/no-Go Decision to Conduct Data Preparation PMI-CPMAI Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A retail company is building a product-recommendation model. The training data includes purchase history from the company's loyalty program, which has 2 million members. However, the loyalty program is only available to customers who spend more than $100 per year. The model will be used to recommend products to all website visitors, including non-loyalty customers. The team has limited budget and time. What is the most appropriate action?

   • AUse the loyalty data but add a demographic re-weighting scheme to approximate the general population.
   • BProceed with training on loyalty data because it is the only available data and the model can still learn general patterns.
   • CTrain the model on loyalty data and restrict the model's deployment to loyalty members only.
   • DCollect a small sample of non-loyalty customer data to assess the gap and then decide.
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A logistics company is preparing a delivery-time prediction model. The final data quality assessment reveals that 5% of the 'distance' values are missing, and the 'weather' column has 3% of values recorded as 'sunny' but the timestamps show those records were from rainy days. The team must decide whether to proceed to data preparation. What should the team do?

   • AProceed and let the model handle the inconsistencies during training.
   • BDrop the weather column entirely because it contains errors.
   • CProceed because 5% missing data is acceptable and weather is a minor feature.
   • DFix the weather inconsistencies and impute the missing distance values before proceeding.
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   When validating that a one-hot encoding transformation was correctly applied to a categorical feature with 5 distinct categories, what should be observed in the output?

   • A5 new binary columns, each indicating presence of a category
   • B4 new columns, each with continuous values
   • CA single column with values 1 to 5
   • DThe original column replaced by a single numeric column
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A data engineer is validating a preprocessing pipeline for a house-price prediction model. The pipeline applies one-hot encoding to 'neighborhood' and standard scaling to 'square_feet'. After running the pipeline, the engineer checks the output and finds that the 'neighborhood' column still contains string values instead of binary columns. What is the most likely cause and the correct next step?

   • AThe one-hot encoding step was skipped or failed; re-run the pipeline and verify the output.
   • BThe string values are acceptable because the model can handle them; proceed with training.
   • CThe standard scaling step corrupted the neighborhood column; re-run only the scaling step.
   • DThe neighborhood column contains too many unique values for one-hot encoding; use label encoding instead.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   What is the primary purpose of documenting data quality findings?

   • ATo increase the speed of model training
   • BTo provide a record for future reference and decision-making
   • CTo replace the need for data quality assessments
   • DTo automatically fix data quality issues
   Reveal Solution Discussion Select an answer first