Train a customer-support model.
Do not miss regions, languages, or devices used in production.
Choose two checks.
Compare user, region, language, device, and seasonal distributions with production and collect missing groups with consent, quality, and label controls.
Detailed explanation
Distribution and use conditions can change.
Coverage improves without ignoring governance.
Their performance and impact disappear.
Training data can be checked against expected use.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 4.1のデータ代表性、公平性、品質を確認する。
Official references
What to confirm
量の多さと本番を代表することを区別し、欠けた集団を確認できる。
Key points
- Representativeness
- Distribution
- Label
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
Representativeness
How well a dataset reflects the people, conditions, and cases the system is intended to serve.
Distribution shift
A difference between the statistical patterns in reference data and current operating data.