Logo Lanfrica

oogunjobi7973-byte/maternal-mortality-prediction

Domaine:

healthcare

Type de record:

model
Créateur:
oog
Hôte:
Early Maternal Mortality Prediction System for Rural Healthcare Systems in Nigeria # Early Maternal Mortality Prediction System An explainable machine learning clinical decision support prototype for maternal mortality risk prediction in rural Nigerian healthcare settings. ## Technology stack - Streamlit frontend - FastAPI backend - Explainable Boosting Machine (EBM) - SHAP explainability - PostgreSQL database - JWT-based basic authentication - Render deployment ## Database The application uses PostgreSQL for persistent storage. On startup, SQLAlchemy creates the required tables if they do not already exist: - `users` - `patients` - `predictions` - `shap_explanations` Prediction requests are associated with the authenticated healthcare worker. Patient assessment records, prediction results, and SHAP feature contributions are persisted in the database. ## Authentication The prototype provides basic healthcare-worker account registration and login. Passwords are stored as secure password hashes, while authenticated API requests use short-lived JWT access tokens. ## Environment variables Backend: - `DATABASE_URL` - Render PostgreSQL connection string - `JWT_SECRET` - long random secret used to sign access tokens Frontend: - `API_BASE_URL` - public URL of the deployed FastAPI backend See `.env.example` for the expected format.