GitHub Forecasting Service
A forecasting app that compares three models to predict GitHub repo growth.
Tech Stack
LSTM Prophet React Flask Docker Python
Overview
This project predicts GitHub repo growth with three different forecasting models. Instead of betting on one approach, it lets you compare the outputs directly.
Architecture
- LSTM for shorter-term sequence-heavy forecasting.
- Prophet for stronger seasonality handling.
- Statsmodels as a simpler statistical baseline.
Each model runs as its own service in Docker, and the React frontend puts the predictions next to each other.
Why it matters
Different repos behave differently. A side-by-side setup made it easier to see where each model was useful instead of pretending one model solved everything.
Result
The combined system reached about 85% accuracy, with LSTM doing better on short-term patterns and Prophet doing better where seasonality mattered.