Heart Disease Risk (Practice) — Classic ML

⚠️ Disclaimer: This is a practice/portfolio project based on a small, noisy dataset. Results are not for medical use.

What's inside

• Practice_classic_ML.ipynb — single Jupyter notebook with EDA → preprocessing → modeling (classic ML classifiers).
• requirements_from_notebook.txt — quick list of Python deps auto-extracted from the notebook.

Reproduce

python -m venv .venv && . .venv/bin/activate  # (or on Windows: .venv\Scripts\activate)
pip install -r requirements_from_notebook.txt
jupyter lab

Open the notebook and run cells top-to-bottom. Set RANDOM_STATE for reproducibility.

Data

• Not included by default. Either:
  1. Use a public dataset (e.g., UCI Heart Disease) and set DATA_PATH accordingly; or
  2. Place your own CSV under data/raw/heart.csv and do not commit it if it has sensitive info.

If you use public data, add a link in this README and a short Data Card: source, size, license, known issues.

Model Card (Short)

• Task: binary classification (heart disease risk proxy).
• Models: LogisticRegression, RandomForest, XGBoost/CatBoost (depending on availability).
• Metrics: ROC-AUC, PR-AUC, F1 (macro), recall@k (if applicable) via CV.
• Limitations: tiny sample → high variance; potential label noise; not for clinical decisions.

License

• Code: MIT (suggested).
• Data: follow upstream license; do not commit private/PHI data.