Databricks Machine Learning Associate Question 93
Select 3You are tasked with scoring a machine learning model using a feature store table in Databricks. The feature store table contains pre-computed features for your model. Which of the following steps are required to correctly score the model using the feature store?
- A
Retrieve the feature store table as a Spark DataFrame and pass it directly to the model's
predictmethod. - B
Use the feature store API to fetch the features required by the model based on the lookup keys.
- C
Ensure that the feature store table is registered in the Databricks Feature Store.
- D
Join the feature store table to your input DataFrame containing the lookup keys before scoring.
- E
Train the model using the feature store table before scoring.
Show answer and explanation
Correct answers: B, C, D
Explanation
Scoring a model using features from a feature store table involves three key steps: ensuring the feature store table is registered, fetching the required features using the feature store API, and joining the features with the input DataFrame using lookup keys. These steps ensure proper alignment between the features and the model's requirements, enabling accurate scoring.
- A. Incorrect.
Incorrect: While you can retrieve the feature store table as a Spark DataFrame, directly passing it to the model's
predictmethod will not work unless the DataFrame has been processed to align with the model's feature requirements. The feature store API ensures this alignment. - B. Correct.
Correct: The feature store API is specifically designed to fetch only the features required by the model based on lookup keys, ensuring proper alignment and avoiding errors.
- C. Correct.
Correct: For a feature store table to be usable, it must be registered in the Databricks Feature Store. Registration is necessary for structured access and governance.
- D. Correct.
Correct: Joining the feature store table to your input DataFrame using lookup keys is an essential step to align the features with your scoring data.
- E. Incorrect.
Incorrect: Training the model using the feature store table is a separate task and not related to scoring. Scoring requires fetching features and applying the model to the input data.