Databricks Certified Machine Learning Professional - 340 Questions

1. A data scientist is building a SparkML pipeline that includes StringIndexer, OneHotEncoder, VectorAssembler, and LogisticRegression. The pipeline must be reusable for both training and scoring. Which approach correctly constructs this pipeline?

2. A team building a recommendation engine needs to handle implicit feedback data using collaborative filtering in SparkML. Which configuration of the ALS estimator is most appropriate for implicit feedback?

3. A data scientist needs to encode a categorical column with over 1,000 unique values for use in a SparkML model. Which approach avoids creating an excessively wide feature vector while preserving categorical information?

4. A company's ML platform team wants to standardize feature scaling across all numeric features before training a LogisticRegression model. The features have different units and ranges. Which SparkML transformer should they use after VectorAssembler?

5. A data scientist wants to perform hyperparameter tuning for a GBTClassifier using cross-validation in SparkML. They need to search over maxDepth values of [3, 5, 7] and maxIter values of [50, 100, 200]. How many total models will be trained with 5-fold cross-validation?