AI Models, Frameworks, Strategies, and Use Cases AAIR Practice Questions (Page 7)

Questions 31–32

1. 31expert · hard

   A logistics company has deployed a machine learning model to predict delivery delays. Six months after deployment, the model's performance has degraded because of seasonal changes in delivery routes. The company needs to maintain model accuracy while minimizing operational disruption. What is the most appropriate action?

   • AImplement a continuous monitoring system and trigger retraining when performance drops below a threshold
   • BFreeze the model and rely on manual adjustments to the delivery routes
   • CRetrain the model from scratch using all historical data including the new seasonal patterns
   • DReplace the model with a simpler rule-based system that can be easily updated
   Reveal Solution Discussion Select an answer first
2. 32application · medium

   A hospital is evaluating an AI system to flag potential sepsis in ICU patients from electronic health records. The system must provide a confidence score and highlight which vital signs contributed to the alert so clinicians can verify the reasoning. Which AI model category would best satisfy these requirements?

   • AA generative model that synthesizes synthetic patient records to augment the training data
   • BA deep learning model such as a recurrent neural network trained on time-series vital signs
   • CAn interpretable machine learning model such as gradient-boosted trees with SHAP values
   • DA classical rule-based expert system encoding clinical guidelines for sepsis
   Reveal Solution Discussion Select an answer first