Detailed Answer in Step-by-Step Solution:
Objective: Address declining model performance and improve churn prediction.
Analyze Issue: Poor predictions suggest data drift or model staleness—common ML challenges.
Evaluate Options:
A . Drift monitoring: Tracks changes in data distribution—identifies root cause of accuracy drop—
correct.
B . Redeploy the model: Repeats deployment without fixing the issue—ineffective alone.
C . Operational monitoring: Tracks system health (e.g., latency), not prediction quality.
D . Retrain the model: Updates model with new data—directly improves accuracy—correct.
E . Validate with recent data: Checks performance but doesn’t fix—diagnostic, not corrective. Reasoning:
A diagnoses drift (cause), D retrains to adapt (solution)—best combo to verify and lower churn.
Conclusion: A and D are correct.
OCI documentation advises: “Drift monitoring (A) detects shifts in data distribution that degrade model
performance, while retraining (D) with fresh data restores accuracy.” Redeployment (B) doesn’t address
drift, operational monitoring (C) focuses on infra, and validation (E) is a check—not a fix. A and D align
with OCI’s model maintenance strategy.
: Oracle Cloud Infrastructure Data Science Documentation, "Model Monitoring and Retraining".
