MLOps Best Practices MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A telecom company has a customer churn model that is retrained monthly. The team wants to compare the performance of the new model against the current production model before deciding to deploy. What is the most appropriate way to do this?

   • AAsk the data scientist to choose the better model based on intuition.
   • BCompare the new model's accuracy on the training data with the old model's accuracy on the same data.
   • CDeploy the new model to production immediately and monitor for a week.
   • DRun an A/B test in production, sending a small percentage of traffic to the new model.
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A retail company has a demand forecasting model that is retrained weekly. The team is experiencing issues with model performance in production, and they suspect it is due to data drift. They have implemented monitoring but are overwhelmed by alerts. What is the best way to improve their monitoring?

   • ASet appropriate thresholds for alerts and prioritize monitoring on the most important features.
   • BDisable monitoring and rely on manual checks.
   • CIgnore alerts that are not related to accuracy.
   • DMonitor every feature with the same threshold.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A technology company is planning to implement MLOps for its recommendation system. The team has both data scientists and software engineers. They want to establish a clear process for model development, deployment, and monitoring. What is the most important element of their MLOps strategy?

   • ALet each team work independently without coordination.
   • BFocus only on model development and ignore deployment and monitoring.
   • CUse the same tools for both data science and engineering without any customization.
   • DDefine clear roles and responsibilities for each stage of the ML lifecycle.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A data science team at a retail company has been manually retraining a churn prediction model every quarter. The process involves a data scientist exporting training data, running a notebook, and manually uploading the resulting model to the model registry. The team wants to reduce human error and ensure that every retraining run uses the same data version and code. Which approach best aligns with MLOps best practices?

   • AContinue the manual process but add a checklist to ensure the data scientist follows the same steps each quarter.
   • BHave the data scientist use a different notebook each time to incorporate the latest data science techniques.
   • CCreate a scheduled Databricks job that runs the training notebook and automatically registers the model with a new version in the registry.
   • DStore the model in a shared folder on the cluster instead of the model registry.
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A manufacturing company is deploying a predictive maintenance model. The operations team requires that the model's predictions be reproducible for audit purposes. The data science team uses MLflow for tracking. What is the most important practice to ensure reproducibility?

   • ASave only the final model artifact without any metadata.
   • BLog the exact code, data version, and hyperparameters for each training run in MLflow.
   • CStore the training notebook in a shared drive without version control.
   • DUse a random seed that changes each run to explore different models.
   Reveal Solution Discussion Select an answer first