Module 07 Assessment: Pipeline Reproducibility Check
Assessment ID: ML-M07-QA01 Estimated active time: 35-50 minutes Status: Draft
Part A: Concept checks
Answer in one or two sentences.
- At what point in your pipeline is the split performed, and why there?
- Which statistics are computed from training data only?
- What score do you get if the scaler is fitted on the whole dataset instead?
- Which of those two numbers is honest, and why?
- What does your pipeline do with a category it has never seen?
Part B: Applied task
Use the supplied synthetic dataset to complete the module activity: Build a pipeline with imputation, scaling, one-hot encoding, and a model.
Part C: Explanation
Explain the leak you demonstrated, why it inflates the score, and what the pipeline still does not handle.
Rubric
| Level | Evidence |
|---|---|
| Pass | Completes the activity, explains the output in plain English, compares or limits the result properly, and avoids overclaiming. |
| Revise | Completes most of the task but misses one important comparison, limitation, or data-safety boundary. |
| Not yet | Treats synthetic results as real-world proof, omits the required evidence, or ignores the module safety rule. |
Safety rule
Do not use real personal, confidential, employer, client, health, financial, authentication, or sensitive data.
