Objective 4.3 Maintain the Model Postproduction AIP-210 Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A credit union deploys a model that approves small personal loans. Six months after launch, the approval rate has dropped from 62% to 48%, but the model's accuracy on the original test set remains high. The data science team suspects the change is due to shifts in applicant demographics and economic conditions. What should the team do FIRST to confirm this suspicion?

   • ARoll back to the previous model version to restore the original approval rate.
   • BCompare the distribution of current input features against the training data distribution using drift detection metrics.
   • CReview the model's documentation to check if the approval criteria were correctly implemented.
   • DRetrain the model immediately with the most recent three months of loan application data.
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A research institute develops a model to predict disease outbreaks. The model is used by public health officials. The institute receives a request from another country to use the model. The model was developed with data from the institute's country, and the data may not be representative of the other country. The institute wants to share the model but is concerned about ethical and practical issues. What should the institute do?

   • ADocument the model's limitations and provide guidance on validation before use in a new context.
   • BShare the model without any documentation to encourage adoption.
   • CRefuse to share the model because it may not work in the other country.
   • DRetrain the model with data from the other country before sharing.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   Which activity is a key component of model governance?

   • AAutomatically approving all model deployments
   • BMaintaining an audit trail of model changes and decisions
   • CDeleting old models to reduce storage costs
   • DKeeping model documentation private to the data science team
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A financial institution must provide an audit trail for a credit scoring model that was retired last year. Regulators ask for the model's training data, hyperparameters, and the rationale for its retirement. The team discovers that this information was never recorded. What should the team have done to avoid this situation?

   • AIncluded the model's code in the company's source control repository.
   • BMaintained a model inventory with metadata and retirement documentation throughout the model's lifecycle.
   • CCreated a backup of the production database at the time of retirement.
   • DStored the model binary in a shared network drive for easy access.
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A healthcare analytics firm maintains multiple versions of a patient readmission prediction model. A new version performed better on validation but caused a significant drop in performance in production. The team needs to revert to the previous version quickly while preserving the ability to investigate what went wrong. What should the team do?

   • ADelete the new version and all its associated data to prevent accidental reuse.
   • BRetrain the new version with more data and redeploy it immediately.
   • CKeep the new version in production and monitor it for another week.
   • DRoll back to the previous model version and keep the new version's artifacts and metadata for analysis.
   Reveal Solution Discussion Select an answer first