Question 15

Explanation

Correct Answer:
✅ To represent features from Feature Store for data augmentation.

Explanation:

The feature_store_lookups parameter in Databricks AutoML is used to augment datasets with additional features stored in the Databricks Feature Store.

• Why is this useful?

  • In machine learning, using enriched feature sets often improves model performance.

  • Instead of manually joining raw datasets with feature tables, AutoML automatically retrieves relevant features from the Feature Store using feature_store_lookups.

  • Ensures that feature engineering steps are consistent and reproducible across ML workflows.

Example: Using feature_store_lookups in Databricks AutoML

import databricks.automl
from databricks.feature_store import FeatureLookup
 
# Define feature lookups from Databricks Feature Store
feature_store_lookups = [
    FeatureLookup(
        table_name="customer_features",
        lookup_key="customer_id",
        feature_names=["age", "income", "customer_segment"]
    )
]
 
# Run AutoML with Feature Store augmentation
databricks.automl.classify(
    dataset=df, 
    target_col="churn_label",
    feature_store_lookups=feature_store_lookups
)

• feature_store_lookups ensures that additional customer features (age, income, customer_segment) are automatically joined with df before training.

• This enables better model accuracy without manual feature engineering.

Why Other Options Are Incorrect?

1. "To control the duration of an AutoML run."

   • Incorrect, because timeout_minutes is used for controlling execution time, not feature_store_lookups.

2. "To list algorithm frameworks that AutoML should not consider."

   • Incorrect, because excluded_algorithms controls which models (e.g., xgboost, random_forest) are skipped, not feature_store_lookups.

3. "To specify the path to the directory in the workspace."

   • Incorrect, because workspace paths are set using the experiment_dir parameter, not feature_store_lookups.

Final Answer:

✅ feature_store_lookups is used in Databricks AutoML to represent features from the Feature Store for data augmentation.