Activity: Create a distance-sensitive model report
Estimated active time: 45-75 minutes
Scenario
Use the synthetic learner-support dataset. The fictional platform wants to decide which model families deserve a place in an algorithm portfolio.
Task
Fit k-nearest neighbours at several values of k, then repeat without scaling, and report what distance is actually measuring in each case.
Required sections
- Problem and decision.
- Validation score at three values of k.
- The same comparison with unscaled features, and the size of the difference.
- Which feature dominates the distance when scaling is omitted, and why.
- Keep, delay, or reject against the Module 1 baselines.
- Limitation note.
Rules
- Scale before computing any distance. Report the unscaled result only as a contrast.
- State how k trades bias against variance in your results, not in the abstract.
- Do not use the test set to choose k.
Output
Use the relevant learner-resource template from learner-resources/ or write a one-page memo with the same sections.