ML Model Evaluation with Task-Correct Metrics
Design a model-evaluation methodology that picks task-correct metrics and addresses imbalance, generalization gap, and bias, interpreting results, without inventing metrics or unmeasured numbers.
Plus
Model: claude-sonnet-4-7
2,055 previews
0 uses
Knowledge: 2026-Q2
This prompt requires the plus plan
Upgrade plan
Your current plan: free
Evaluating an imbalanced fraud-detection model before production, choosing correct metrics for a diagnostic model, preparing an auditable evaluation report for a model-risk board.
#تقييم نماذج#مقاييس#تحيّز#تعميم
Built-in quality guards
Anti-hallucination
No fabricated metrics, functions, or unmeasured performance numbers
Security Check
Avoid printing individual records and respect PDPL in sensitive-group analysis
Completeness
Cover metrics, protocol, error analysis, generalization, and fairness without truncation
Format Check
Adherence to the six-section structure, results table, and decision recommendation