Live Demo: econ-pulse-9cuc.vercel.app
A student-centered economic dashboard that converts public FRED macroeconomic data into practical financial awareness tools for students, young adults, and early investors.
Built for HackDavis 2026, EconPulse combines data pipelines, machine learning models, scenario simulation, and an interactive frontend to make economic conditions easier to interpret.
Economic indicators are often scattered across government dashboards, dense reports, and disconnected datasets. EconPulse brings these indicators into one interface and translates them into a more intuitive view of economic health.
The project focuses on questions like:
- Is the labor market strengthening or weakening?
- Is inflation pressure rising or cooling?
- How do different indicators combine into an overall economic health score?
- How do model predictions change under different economic scenarios?
- Economic dashboard using FRED-style macro indicators
- Processed data pipeline with cached/mock fallback support
- Model Hub comparing Multiple Linear Regression, XGBoost, LightGBM, and an MLP Neural Network- Scenario simulator for changing category assumptions
- Interactive frontend built with React, TypeScript, Vite, and Recharts
- Backend pipeline using Python, pandas, NumPy, and scikit-learn
Frontend
- React
- TypeScript
- Vite
- Recharts
- lucide-react
- CSS
Backend
- Python
- pandas
- NumPy
- scikit-learn
- Optional XGBoost
EconPulse uses public macroeconomic indicators from FRED where available, with cached/mock fallback data for development and demo stability.
Data
- FRED API when available
- Cached or mock data fallback
- Processed JSON files served to the frontend
EconPulse/
├── backend/
├── data/
├── frontend/
│ └── public/data/
└── README.md
python -m backend.run_pipelineThe backend writes processed files to data/processed and syncs frontend-ready outputs into:
frontend/public/data
cd frontend
npm install
npm run dev| Section | Purpose |
|---|---|
| Dashboard | Displays current economic health and category-level scores |
| Model Hub | Compares model forecasts and validation metrics |
| What-If Simulator | Tests scenario changes with slider-based inputs |
| Documentation | Explains pipeline, stack, and data files |
The model hub forecasts overall economic health three months ahead. The project compares multiple supervised learning models so users can see how different model assumptions affect predictions.
Current model comparison includes:
- Multiple Linear Regression
- XGBoost
- LightGBM
- MLP Neural Network
The Model Hub compares several supervised learning models for a 3-month economic health forecast.
| Model | Forecasted Health | MAE | RMSE | R² |
|---|---|---|---|---|
| Multiple Linear Regression | 41.8 | 2.72 | 3.29 | 0.871 |
| XGBoost | 41.6 | 2.06 | 2.51 | 0.925 |
| LightGBM | 42.1 | 2.08 | 2.41 | 0.930 |
| MLP Neural Network | 39.7 | 2.38 | 2.84 | 0.903 |
Lower MAE and RMSE indicate smaller forecast errors. Higher R² indicates that the model explains more variation in the future economic health target.
LightGBM performed best overall, with the lowest RMSE and highest R² among the tested models. XGBoost had the lowest MAE, making both gradient-boosting models stronger than the linear and neural-network baselines.
This project helped me connect economics, statistics, and software engineering in one system. I practiced building an end-to-end data product, from data ingestion and preprocessing to modeling, frontend design, and deployment.
- Add more economic indicators and metadata
- Improve model validation and backtesting
- Add time-series models
- Add confidence intervals or uncertainty bands
- Improve responsive design
- Add richer documentation for each indicator
- Add automated tests for backend data transformations
EconPulse is an educational project and is not financial advice.
EconPulse reflects my interest in quantitative finance, macroeconomic data, and applied machine learning. I built this project to show that I can work across data pipelines, statistical modeling, and frontend product design.