🩸 Global Blood Donation Registry — Decision-Grade Outreach Policy

A production-style notebook that converts a donor registry into calibrated outreach scores and budgeted outreach decisions.

Notebook: global-blood-donation-registry.ipynb
Case study: CASE_STUDY.md

---

What this repository includes

• A single notebook: global-blood-donation-registry.ipynb
• Exported artifacts under ./artifacts/ (scored donors + model bundle + metrics)

---

Core output

A calibrated probability for eligible donors:

Outreach Score = P(donated_next_6m)

Then the notebook turns scores into actions:

• Top‑K outreach for a a set outreach capacity
• Net‑benefit threshold that maximizes expected value under simple costs/benefits

---

Dataset

The notebook expects these CSV files:

Required

• blood_donation_registry_ml_ready.csv
• blood_population_distribution.csv
• blood_compatibility_lookup.csv
• data_dictionary.csv

Local (recommended)

1. Download the dataset files.
2. Place them under: data/raw/

See data/raw/README.md.

Kaggle

If local files are not present, the notebook falls back to the Kaggle input paths.

---

Getting started

1) Install

python -m venv .venv
# Windows: .\.venv\Scripts\activate
# macOS/Linux: source .venv/bin/activate

pip install -r requirements.txt

2) Run the notebook

Open and run:

• global-blood-donation-registry.ipynb

The notebook will export:

• artifacts/donor_outreach_scored.csv
• artifacts/model_bundle.joblib
• artifacts/metrics_holdout.json

See artifacts/README.md.

---

Methodology notes

• Scores are computed only for eligible donors.
• Calibration makes thresholds usable for decision policies.
• Threshold policies make trade-offs explicit (contact cost vs expected donation benefit).

---

License

MIT (code). Dataset licensing depends on the dataset source you download.