Operationalize and Monitor Models SNOWPRO-ADVANCED-DATA-SCIENTIST Practice Questions (Page 6)

Questions 26–27

1. 26application · medium

   A financial services firm has a fraud detection model in production. The model is retrained monthly using the latest labeled data. Recently, the model's precision has declined while recall has remained stable. The data science team suspects that the fraud patterns have evolved. What is the most appropriate next step?

   • AImmediately roll back to the previous model version and keep it in production indefinitely.
   • BIncrease the alert threshold for precision so that the model does not trigger unnecessary alerts.
   • CSwitch to a different model algorithm immediately to improve precision.
   • DInvestigate feature distributions and model performance on recent data to confirm drift, then retrain with more recent data or adjust the retraining frequency.
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   Which step is typically the FIRST action in an incident response process for a model anomaly?

   • AInvestigating the alert to confirm and understand the anomaly
   • BRolling back the model to a previous version
   • CDeleting the model from production
   • DImmediately retraining the model with new data
   Reveal Solution Discussion Select an answer first