Client project · Early-stage healthcare startup · 6-person Agile team · Jan–May 2024
Hospital Risk Prediction Web App
A machine learning dashboard that predicts hospital stress levels from simulated operational data.
My role: Designed the MySQL schema and built the background service that polls simulated hospital systems.
- Python
- Django REST
- Scikit-learn
- MySQL
- Docker
Developed a Scikit-learn machine learning model to predict hospital stress levels using synthetic data.
Built a Python data generator and a Django REST API that simulate polling hospital systems and persist collected data in MySQL.
Integrated user authentication into a Bootstrap-based dashboard for viewing real-time risk-level predictions.
Collaborated on a six-person Agile team, delivering functional code and documentation to the client in biweekly sprints.
How it works
Simulated hospital systems
A dummy API in its own Docker container serves staffing and bed-occupancy readings.
Polling service
Built by meAn APScheduler background job requests a new reading every 10 seconds and validates it with a Django REST serializer before saving.
MySQL database
Built by meReadings, training data, and predictions live in separate tables; risk predictions and incident logs reference the reading they belong to.
- hospital_data
- training_data
- risk_data
- incident_log
- email_recipient
Prediction job
A Scikit-learn random forest trained on synthetic data scores each new reading and stores the risk level and probability.
Dashboard
Built by meAn authenticated Bootstrap dashboard shows the latest risk level. I designed its layout and the prediction display.
- GitHub repository
- No longer deployed: it ran on a client-hosted database that has since been retired.