Activity: Create a pruned tree 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 decision trees at several depths, extract the rules from the best shallow tree, and report where interpretability stops being real.

Required sections

1. Problem and decision.
2. Train and validation score at depths 1, 3, 5 and unrestricted.
3. The printed rules from the depth-3 tree, in plain English.
4. The depth at which validation stops improving while training continues.
5. Keep, delay, or reject against the Module 1 baselines.
6. Limitation note.

Rules

• Report train and validation together. A tree's training score is near-meaningless alone.
• An unrestricted tree is not interpretable, whatever the marketing says. Say so if it wins.
• Do not prune against the test set.

Output

Use the relevant learner-resource template from learner-resources/ or write a one-page memo with the same sections.