BRACKEN FRAUD LEARNING LAB
Mickai Training Academy

Educational only. Mickai Training Academy is not an accredited training provider.
This invented offline exercise gives no professional qualification, investigation authority,
financial or legal advice, or permission to contact customers or operate accounts.

Start with learner_tasks.md and investigator_form.md. Use the tables in the course for
manual work, or Python 3.10 or later (standard library only):
  python checker.py
  python -m unittest -v test_checker.py
Run from this extracted folder. No account, network, model, external package or credential
is required. checker.py prints calculations; it does not write files, close cases or send
messages. Tests use copies in memory and a temporary folder, preserving supplied inputs.

alerts.csv: 500 distinct fictional alerts and fixed Q1 indices.
baseline_outcomes.csv: complete later labels for the baseline, available 12 November.
shifted_outcomes.csv: a separate hypothetical pattern shift, not baseline corrections.
prior_queue.csv: 20 distinct older cases with an already-breached invented ageing trigger.
feedback_candidates.csv: six candidate labels, four deliberately ineligible at cutoff.
packet.json: exact dates, policy assumptions and three source case cards.
answers.md: complete worked activity and chapter-review answers.
expected_results.json: independently specified arithmetic ledger.

Scores are uncalibrated supplied priority indices, not probabilities. Labels, authority
flags and policy thresholds are declared fictional inputs. The checker cannot authenticate
an investigator or decide whether a real investigation was justified. Read the printed
course for methodological limits and scoped primary-source references. Preserve the
original files and use labelled copies if exploring changes. Never replace missing labels
with zero. Practical work is ungraded; a separate knowledge exam does not certify it.
