lesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and trackerlesson 5 of 5 · mission: feature pipeline and tracker
Checkpoint
One last thing before we move on. pass this to mark the lesson done, or skip and keep moving.
The whole loop, miniature: (1) run each raw row through features() and the contract — the -2.0 tickets row gets quarantined; print f"rows kept={k} quarantined={q}". (2) Sweep cutoffs (2, 4, 6) over the CLEAN data, logging each run with config {"cutoff": ..., "features": "v1"} and printing f"run cutoff={c}: acc={a}". (3) Register the best run (max acc) as churn v1 production with its config and data provenance ("raw_rows_v1"), printing f"registered: churn v1 production acc={a} config={cfg}". Raw rows in — contract, log, registry out.