Insights about Python packages.
The Commonly used together section comes from scanning all of
GitHub for files named
requirements.txt, and building a
recommender system based on matrix factorization.
Download counts are based on server logs of PyPI for each package,
published by PSF.
Both of these datasets are published on BigQuery, which made compiling
the information extremely easy.
I did all my data exploration and model building in
Deepnote, which is a lovely new way
to work with data science notebooks, with real-time collaboration and