Automated Retraining MACHINE-LEARNING-PROFESSIONAL Practice Questions (Page 5)

Questions 21–25

1. 21expert · hard · select all that apply

   A company has a model in production that predicts customer churn. They want to set up automated retraining triggers. Which of the following are valid triggers for automated retraining? Select all that apply.

   • AData drift detection: trigger retraining when the distribution of input features shifts significantly.
   • BScheduled retraining: trigger retraining on a fixed schedule (e.g., weekly) regardless of drift or performance.
   • CManual trigger: a data scientist manually starts the retraining job when they think it's needed.
   • DNew data availability: trigger retraining whenever new labeled data arrives.
   • EPerformance degradation: trigger retraining when the model's accuracy or other metrics decline below a threshold.
   Check answers Discussion Select an answer first
2. 22application · medium

   A credit risk team has a model in production that predicts loan default probability. They want to automatically retrain the model when its AUC (area under the ROC curve) on recent data drops by more than 0.05 compared to the baseline. They have a monitoring job that computes AUC on a rolling window of new labeled data. What is the best way to implement the retraining trigger?

   • AHave the monitoring job check if the AUC drop exceeds the threshold, and if so, trigger the retraining pipeline via the Databricks Jobs API.
   • BSet up an alert in the Databricks SQL dashboard that notifies the team when AUC drops, and let them manually start the retraining job.
   • CUse a Delta Live Tables pipeline to continuously retrain the model whenever new labeled data arrives.
   • DCreate a Databricks job that runs the retraining pipeline every week, regardless of AUC.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   What is the primary purpose of integrating data drift detection into an automated retraining pipeline?

   • ATo improve the speed of model inference at runtime
   • BTo reduce the storage cost of the training dataset
   • CTo automatically initiate retraining when the input data distribution shifts significantly from the training data
   • DTo encrypt the data before it is sent to the model
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   What is the primary purpose of using a model registry in an automated retraining workflow?

   • ATo store the raw training data for future use
   • BTo manage and version retrained models for tracking, selection, and deployment
   • CTo monitor the model's runtime performance
   • DTo automate the retraining trigger detection
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A retail company uses a Databricks ML pipeline to predict daily store demand. They want to automatically retrain the model when the distribution of the input features (e.g., promotions, holidays) shifts significantly. They have a monitoring job that computes a drift metric on the latest batch of data. What is the most efficient way to trigger retraining?

   • AConfigure a Databricks job that runs the retraining pipeline on a fixed schedule (e.g., daily) regardless of drift.
   • BUse the Databricks Jobs API to programmatically trigger the retraining pipeline when the drift metric exceeds a threshold.
   • CManually review the drift metric weekly and decide whether to run the retraining pipeline.
   • DSet up a Delta Live Tables pipeline that continuously retrains the model on every new batch of data.
   Reveal Solution Discussion Select an answer first