Data Analytics / Data Science Capstones

This repository contains mini and full capstones for Practera data science modules.

Launch in Browser

Open the interactive JupyterLite environment (no install required):

Launch JupyterLite

Everything runs in your browser via Pyodide — no server, no setup, no accounts.

Notebooks

Practera Capstones

Capstone	Notebook	Direct Link
Exploratory Data Analysis	practera/eda/task.ipynb	Open
Data Cleaning	practera/cleaning/task.ipynb	Open
Clustering	practera/clustering/task.ipynb	Open
Regression	practera/regression/task.ipynb	Open

SkillsBuild Sustainability

Capstone	Notebook	Direct Link
NYC Water Project 1	skillsbuild/sustainability/nyc_water_project_1.ipynb	Open
NYC Water Project 2	skillsbuild/sustainability/nyc_water_project_2.ipynb	Open

How It Works

This repo uses JupyterLite to serve a full Jupyter environment as a static site on GitHub Pages. On every push to trunk, a GitHub Action builds the site and deploys it.

Packages available in the Pyodide kernel include pandas, numpy, matplotlib, scikit-learn, scipy, and statsmodels. Additional pure-Python packages (like plotly) can be installed at runtime via %pip install.

Known Limitations

• Google Sheets notebook (practera/eda/task_gsheets.ipynb) is excluded — it requires server-side Google API credentials which aren't available in a browser environment.
• NYC Water notebooks fetch data from the NYC Open Data API. Pyodide supports HTTP fetch, but pd.read_json(url) may need to use pyodide.http.open_url() instead. The local fallback file project_2_data.json is included.
• Excel files referenced by the cleaning and clustering notebooks (test_data.xlsx, processed_data.xlsx, mainstream_media.xlsx) are not in this repo and were previously provided externally.
• First load takes 10-20 seconds as Pyodide downloads and initializes in your browser. Subsequent visits are faster due to browser caching.

Development

To build the JupyterLite site locally:

pip install -r requirements.txt
mkdir -p content
cp -r practera content/
cp -r skillsbuild content/
jupyter lite build --contents content --output-dir dist

Then serve dist/ with any static file server:

python -m http.server -d dist 8000

Previous Setup

This repo previously used BinderHub + nbgitpuller via intersective/binder-base. That approach has been replaced with JupyterLite for a lighter-weight, zero-infrastructure solution.