Databricks Machine Learning Professional Question 129
Single answerA company wants to use a machine learning model to predict customer churn for its subscription service. The model is designed to process a large dataset of customer behavior and generate churn predictions for all users once per week. Which deployment approach is most appropriate for this use case?
- A
Batch deployment, where predictions are generated periodically for all customers and stored for later use
- B
Online deployment, where predictions are generated in real-time for each customer as they interact with the service
- C
Edge deployment, where the model is deployed on user devices for local prediction
- D
Hybrid deployment, combining batch and real-time predictions for maximum flexibility
Show answer and explanation
Correct answer: A
Explanation
Batch deployment is ideal for scenarios where predictions are required periodically rather than in real-time. In this case, the company needs to generate churn predictions weekly, making batch deployment the most efficient and appropriate solution. It allows the model to process a large dataset at once and store predictions for later use without the overhead of real-time systems.
- A. Correct.
Correct: Batch deployment is the most suitable approach for generating predictions periodically, especially when predictions are not required in real-time. This matches the requirement to process customer data weekly and store predictions for later use.
- B. Incorrect.
Incorrect: Online deployment is unnecessary in this case since predictions are not needed in real-time. Online deployment would add complexity without providing additional value.
- C. Incorrect.
Incorrect: Edge deployment is not relevant here because the predictions are generated centrally for a dataset and not on individual devices.
- D. Incorrect.
Incorrect: Hybrid deployment is overly complex for this use case, as there is no indication of a need for both batch and real-time predictions.