Data Acquisition GMLE Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A data scientist is building a model using data from a public API that provides weather observations. The API sometimes returns missing values for temperature and occasionally returns duplicate records for the same timestamp. The scientist wants to assess data quality during acquisition. Which step should be performed first?

   • AImpute missing temperatures using the mean
   • BRemove all records with missing values
   • CBuild a machine learning model to predict missing values
   • DProfile the data to quantify missing values and duplicates
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A data analyst needs to extract data from a PostgreSQL database that contains a table with millions of rows. The analyst only needs records from the last 30 days. Which SQL query approach is most efficient?

   • ASelect all rows and filter in the application code
   • BUse a LIMIT clause to get the first 1000 rows
   • CCreate a new table with all the data and then query it
   • DUse a WHERE clause with a date filter on an indexed date column
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A data team is building a pipeline to acquire customer data from multiple regional sources. Each source uses a different format for dates and phone numbers. The team must produce a unified dataset for a global model. They also need to be able to trace any transformation back to the original source record. Which approach best balances data quality and provenance?

   • AStandardize all fields during ingestion and store only the transformed data, with a mapping table documenting the transformations.
   • BStore raw data as-is in a landing zone, then create a curated layer with standardized fields, keeping a lineage link between the two.
   • CLoad all data into a single table and manually fix inconsistencies during analysis.
   • DStandardize fields during ingestion and discard the raw data to save storage costs.
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A data science team is developing a model that requires data from a third-party vendor. The vendor updates their dataset daily and does not provide versioning. The team needs to ensure that model training is reproducible and that they can audit which vendor data version was used. Which approach is most effective?

   • ARecord only the date of acquisition in the model metadata.
   • BSnapshot the vendor data into a versioned internal storage system immediately after acquisition, and record the acquisition timestamp and source metadata.
   • CUse the vendor's data directly from their API each time training runs.
   • DStore the vendor data in a local file and overwrite it each day.
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A healthcare analytics team is building a pipeline to ingest patient records from multiple hospitals. Each hospital sends CSV files to an SFTP server. The team needs to ensure that data is complete and consistent before loading into the warehouse. What should the ingestion pipeline do?

   • ARely on the SFTP server to validate the files
   • BLoad all files immediately and fix issues later
   • CValidate each file for schema and required fields, then load only valid files
   • DOnly load files that are larger than a certain size
   Reveal Solution Discussion Select an answer first