Databricks Machine Learning Professional Question 162
Select 3A data science team is tasked with scoring a large batch of incoming data for fraud detection using a pre-trained machine learning model in Databricks. The team decides to use the score_batch operation instead of scoring the data in real-time. What are the practical benefits of using the score_batch operation in this scenario?
- A
It allows scoring large volumes of data more efficiently by processing data in batches.
- B
It reduces the need to maintain an always-on endpoint for real-time inference.
- C
It provides better prediction accuracy compared to real-time scoring methods.
- D
It enables integration with scheduled workflows, such as ETL pipelines, for periodic batch processing.
- E
It automatically tunes hyperparameters for the model before scoring the data.
Show answer and explanation
Correct answers: A, B, D
Explanation
The score_batch operation is ideal for scenarios where large datasets need to be scored efficiently without the need for real-time inference. It reduces operational overhead by not requiring an always-on endpoint and can be seamlessly integrated with scheduled workflows, such as batch ETL pipelines. However, it does not modify the model or improve accuracy, as it relies on a pre-trained model for inference.
- A. Correct.
Correct: The
score_batchoperation is optimized for batch processing, making it more efficient for handling large datasets as opposed to scoring individual records in real-time. - B. Correct.
Correct: Batch scoring does not require an always-on endpoint, which is necessary for real-time scoring. This reduces operational costs and complexity.
- C. Incorrect.
Incorrect: The
score_batchoperation does not improve prediction accuracy; it uses the same pre-trained model as real-time scoring. Accuracy depends on the model, not the scoring method. - D. Correct.
Correct: Batch scoring can be integrated with scheduled workflows (e.g., ETL jobs) to process data periodically, aligning well with batch processing strategies.
- E. Incorrect.
Incorrect: The
score_batchoperation does not involve hyperparameter tuning. It simply applies an existing trained model to the input data.