Which of the following is NOT a key monitoring metric for deployed CPMAI models?

• A. Prediction accuracy drift
• B. Data input distribution drift
• C. System health such as latency and uptime
• D. Training dataset size

Answer: (D)

Monitoring deployed CPMAI models centers on how the model behaves in production and how reliably the service runs. The key signals are drift in prediction accuracy over time, drift in the input data distribution, and system health metrics like latency and uptime. These tell you if the model’s performance is staying within expectations and if the service remains available and responsive.

Training dataset size is tied to how the model was developed rather than how it performs in production. It doesn’t reflect current performance, data or concept drift, or runtime reliability, so it isn’t used as a live monitoring metric for deployed models.