Objective 4.3 Evaluate Models CERTIFIED-DATA-SCIENCE-PRACTITIONER Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A team uses MLflow to track experiments. They have run 50 experiments with different hyperparameters and logged metrics for each. They now want to programmatically identify the run with the highest F1-score on the validation set and register that model in the MLflow Model Registry. Which approach is most efficient?

   • AExport all runs to a CSV file and sort in Excel.
   • BRetrain the model with the best hyperparameters found in the UI.
   • CManually review each run in the MLflow UI and note the best F1-score.
   • DUse the MLflow Tracking API to query runs, sort by the 'f1' metric, and register the best run's model.
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A data scientist wants to share a trained model with another team for deployment. The model was trained using MLflow and logged as an artifact. What is the most reliable way to share the model?

   • ARegister the model in the MLflow Model Registry and share the registered model version.
   • BSend the raw pickle file via email.
   • CPrint the model coefficients and send them.
   • DCopy the model artifact to a shared drive.
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A team is running multiple experiments with different hyperparameters and wants to compare them systematically. They need to log metrics, parameters, and artifacts for each run, and later retrieve the best run for deployment. Which tool is most appropriate for this task?

   • AKubeflow Pipelines
   • BMLflow Tracking
   • CTensorBoard
   • DA shared CSV file
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   What is the primary purpose of MLflow Tracking?

   • ATo record and query experiments, including parameters, metrics, and artifacts
   • BTo deploy models to production with one click
   • CTo visualize the neural network architecture
   • DTo automatically tune hyperparameters without user input
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A data scientist has trained a model and wants to store it along with its metadata (metrics, parameters, and a description) so that other team members can discover and use it. Which approach best satisfies this requirement?

   • AStore the model in a Git repository.
   • BSave the model as a pickle file and upload it to a shared file server.
   • CRegister the model in the MLflow Model Registry with a description and tags.
   • DEmail the model file to the team.
   Reveal Solution Discussion Select an answer first