Objective 4.3 Maintain the Model Postproduction AIP-210 Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   An online payment platform uses a fraud detection model. The model is retrained weekly, but the team notices that the model's precision has been declining while recall has improved. They suspect that the retraining process is introducing a bias toward catching more fraud but with more false positives. The team needs to maintain a balance between precision and recall. What should they do?

   • ARoll back to the previous model version and stop retraining.
   • BRetrain the model less frequently to reduce the impact of noise.
   • CUse a different algorithm that naturally balances precision and recall.
   • DSet a minimum precision threshold in the retraining evaluation and reject model versions that do not meet it.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   An e-commerce company's product recommendation model is retrained every Sunday using the previous week's clickstream data. Recently, the model's performance on a fixed holdout set has been declining, and the team notices that the retraining data includes a large number of bot-generated clicks. What is the most appropriate adjustment to the retraining process?

   • AAdd a data quality filter to remove bot-generated clicks before retraining.
   • BIncrease the retraining frequency to daily to dilute the impact of bot clicks.
   • CSwitch to a more complex model architecture to better handle noisy data.
   • DFreeze the model and stop retraining to prevent further degradation.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   Which information should be included in model documentation?

   • AThe model's training data sources and preprocessing steps
   • BThe marketing strategy for the model's product
   • CThe exact server IP addresses used during training
   • DThe personal credentials of the model developers
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A bank's fraud detection model is updated frequently. The compliance team requires that every model version be linked to the specific training dataset and evaluation results for audit purposes. What practice should the team implement to meet this requirement?

   • AUse a model registry that stores metadata, including dataset version and evaluation metrics, for each model version.
   • BSave each model version as a separate file in a shared folder.
   • CEmail the model file to the compliance team after each update.
   • DRecord the model version in a spreadsheet manually.
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A pharmaceutical company uses a model to predict drug side effects. The model was approved by regulators under specific conditions. The company plans to retire the model because a new, more accurate model is ready. However, the regulator requires that the old model remain available for a year after retirement for post-market surveillance. What should the company do to comply?

   • ADelete the model and all related data to protect patient privacy.
   • BKeep the model running in production until the retention period ends.
   • CRetrain the old model with new data to keep it accurate during the retention period.
   • DArchive the model and its artifacts, and document the retirement and retention period.
   Reveal Solution Discussion Select an answer first