Optezum - ML Research Forecasting Project

Optezum combines research notebooks (SARIMA/XGBoost + Google Trends) with a Next.js web app for interactive forecasting demos.

Repository structure

• notebooks/ - research pipeline (01 to 06) for data prep, SARIMA baseline, XGBoost with trends, and comparison.
• src/ - Python helper scripts (synthetic sales generation and Google Trends fetching).
• frontend-react/ - Next.js app used for UI interaction, CSV ingestion, and visual forecasting flow.
• data/ - raw/processed project datasets and model result artifacts.

Local development

1) Python environment

pip install -r requirements.txt

2) Frontend app

cd frontend-react
npm install
npm run dev

Open http://localhost:3000.

Research vs app runtime

• Notebooks are the primary place where full research model training is performed.
• The frontend currently provides an interactive, CSV-driven demo pipeline and visual analytics experience.

Deployment (Railway)

• Railway build uses root Dockerfile.
• Runtime settings are managed via root railway.json.
• Frontend assets and app source are under frontend-react/.

Privacy and secrets

• Do not commit .env, credentials, API keys, or personal machine files.
• Existing .gitignore already excludes common private/local artifacts such as .venv/, .env, and node_modules/.