About this project
Rebuilt 2026 from a 2020 BYU INTEX class project.
What it is
A unified rebuild of four older repos that all branched from the same BYU senior INTEX project: a COVID-19 GoFundMe analytics tool. The original was a Django backend serving a React 16 frontend, with a success-prediction model running on Azure ML Studio.
Why the rebuild
Microsoft retired Azure ML Studio in 2024, which broke the model endpoint. The React app was on CRA + Bootstrap 4, which is also gone. The backend was duplicated across two near-identical repos. This rebuild collapses everything into one Next.js app with the model re-trained locally and served as JSON weights — no backend needed.
Stack
- Next.js 15 (App Router) + TypeScript + Tailwind 4
- Recharts for visualizations
- Static generation for all ~4,800 campaign detail pages
- Logistic regression trained with scikit-learn; weights as JSON
- Vercel for hosting
Data
~4,800 GoFundMe campaigns scraped during the March 2020 surge, with title, category, goal, raised amount, donator count, hearts, shares, location, and charity verification. The success label is raised >= goal. Note that engagement features (donors, shares) are partial post-hoc signals — useful for explaining what correlates with success, but they leak future information for true forecasting. Treat the predictor as a what-if tool, not a forecast.