Databricks Machine Learning Associate Question 446
Single answerYou are working on a machine learning project in Databricks and have trained a model to predict customer churn. The feature data required for scoring the model is stored in a feature store table. How should you retrieve the features from the feature store table and score your model?
- A
Use the Feature Store API to load the feature table as a DataFrame and join it with the input data before scoring.
- B
Directly query the feature store table using SQL and pass the resulting DataFrame to the model for scoring.
- C
Use the Feature Store API to load the feature table and directly pass it to the model for scoring without any additional steps.
- D
Export the feature store table to a CSV file, load the file into a DataFrame, and use it for scoring.
Show answer and explanation
Correct answer: A
Explanation
The Databricks Feature Store is designed to facilitate the retrieval and integration of features into machine learning workflows. The recommended approach is to use the Feature Store API, which ensures that feature transformations, metadata, and joins with input data are correctly applied before scoring the model. This ensures consistency, reproducibility, and efficiency in the machine learning pipeline.
- A. Correct.
Correct: The Feature Store API in Databricks allows you to retrieve features from a feature store table, ensuring that the feature transformations and joins are applied correctly. This is the best practice for integrating features into your scoring pipeline.
- B. Incorrect.
Incorrect: While you can query the feature table using SQL, this approach does not ensure that feature transformations or metadata associated with the feature store are preserved. It is not the recommended method when using the Databricks Feature Store.
- C. Incorrect.
Incorrect: The Feature Store API cannot directly pass the feature table to the model for scoring without ensuring the features are joined with the input data. This step is necessary for accurate scoring.
- D. Incorrect.
Incorrect: Exporting the feature store table to a CSV file is not a practical or efficient approach in Databricks. It also breaks the integration and traceability provided by the Feature Store.