ML Case Labs — Validation, Boosting, Segmentation, Pipelines
Back to M4 — Pipelines Explainability + Capstones

Capstone A: Campaign Classifier

Classifier + calibration + decision memo. Checks then LLM review.

71 minutesProject submission
🎯 Free Guest Mode: You are learning for free. Sign in to save your completion progress and quiz answers.

Assignment Instructions

Brief

Ship the full campaign classifier: time-split validation, calibrated model, cost-matrix threshold, decision memo. Deterministic checks then LLM review.

Dataset

Full campaign data + hidden holdout window.

Starter guardrails (88)

  • Full pipeline from the m2-l8 skeleton; threshold frozen pre-holdout.
  • Memo template: decision, metric table, calibration evidence, risks.

Deterministic checks

Holdout PR-AUC >= 0.70; Brier <= 0.16; temporal split intact; memo complete.

100-pt rubric

  • Validation discipline: 20
  • Model + calibration quality: 25
  • Threshold economics: 20
  • Holdout metric bar: 15
  • Decision memo: 15
  • Reproducibility: 5
Checking your access…