Databricks Machine Learning Associate Question 437
Select 3You are tasked with training a machine learning model to predict customer churn. You decide to use features from a feature store table in Databricks. Which of the following steps are required to correctly retrieve features from the feature store and train the model?
- A
Use the
FeatureStoreClientto load the feature table as a DataFrame. - B
Join the feature table's DataFrame with the target labels to prepare a training dataset.
- C
Directly pass the feature table's name to the chosen ML model for training.
- D
Perform feature lookups to enrich the feature DataFrame with additional information if needed.
- E
Log the feature table's metadata using the
mlflow.log_artifact()method before training the model.
Show answer and explanation
Correct answers: A, B, D
Explanation
To train a model using features from a feature store table in Databricks, you must retrieve the features using the FeatureStoreClient, combine them with the target labels to create a suitable training dataset, and optionally enrich the features through feature lookups. While logging metadata is a useful practice, it is not a required step for model training.
- A. Correct.
Correct: To interact with a Databricks feature store table, the
FeatureStoreClientis used to retrieve the features as a DataFrame. - B. Correct.
Correct: To train a model, the feature DataFrame must be joined with the target labels to create a training dataset.
- C. Incorrect.
Incorrect: You cannot directly pass the feature table's name to the model. Features must first be retrieved and prepared in a DataFrame format.
- D. Correct.
Correct: Feature lookups can be performed to enrich the primary feature DataFrame with additional relevant feature information.
- E. Incorrect.
Incorrect: Logging metadata is not a required step for training the model with features from a feature store table. Instead, logging is typically done after the model is trained.