FRA-RIG Breast Cancer (sklearn demo): block-weighted risk scoring

Research-only model weights (joblib) and calibration report (JSON) for the RIG–FRA breast cancer risk scorer.

Data: sklearn.load_breast_cancer.

5-fold CV (avg): ACC 0.967 | ROC-AUC 0.989 | PR-AUC 0.990 | F1 0.973.
Youden threshold: 0.643.

Files:

fra_rig_breast.joblib — trained system (prep + weights + calibrated head)

fra_rig_report.json — CV metrics summary

(optional) example inference output (CSV/JSONL)

FRA-RIG-breast-code-v1.0.0.zip — source code (MIT). Mirrors GitHub release v1.0.0.
Quick run: pip install -r requirements.txt → python predict_demo.py
Repo: https://github.com/AdmailFRA/FRA-RIG-breast

License: CC BY 4.0 for model weights and report; MIT for code archive.
Disclaimer: research only, not for clinical or diagnostic use.