Algorithm Comparison Report Template
Use this when comparing model families. Keep it short and evidence-based.
Problem
- Prediction task:
- Target:
- Unit of prediction:
- Decision supported:
Data and split
- Dataset:
- Train/validation/test split:
- Leakage checks completed:
- Important data limits:
Baseline
- Baseline model:
- Baseline metric:
- Why this baseline is fair:
Candidate models
| Candidate | Why try it | Main sensitivity | Metric result | Keep, reject, or inspect more |
|---|---|---|---|---|
Recommendation
Choose the simplest adequate model. If no model is adequate, say so.
Limitation note
Name at least one data, metric, stability, fairness, calibration, or maintenance limit.