NuGet data-ecosystem package popularity (monthly)

Monthly popularity ranking of .NET/NuGet packages in the data/ML ecosystem. About two dozen data-relevant terms (machine learning, dataframe, database, ETL, analytics, visualization, statistics, time series, NLP, ...) are searched against the official keyless NuGet V3 search API (prerelease excluded) into a deduplicated candidate universe (NuGet ids are case-insensitive; every query term that matched is recorded as collection provenance). Packages clearing the 500,000-download floor get their registration metadata: latest listed version, its publication date, license expression and deprecation status (deprecated packages are dropped). Packages are classified into a curated 10-category taxonomy by a deterministic ordered rule set over their tags and description/keywords. The popularity score is the min-max-normalized log10 total downloads (0-100) and the recency score the min-max-normalized release recency (100 for a release <=30 days old, decaying linearly to 0 at 24 months); popularity_score is their 50/50 composite (0-100) and popularity_rank orders by popularity_score descending (ties: total downloads desc, then package name). Columns: month, fetch timestamp, package name / URL / latest version / description / license id / publication date, tags / taxonomy category, matched search terms, exact source API URLs for each row's values (search record, registration index — any number is one GET away from its source), total downloads, days since latest release, download/recency/popularity scores (0-100), popularity rank, deprecated flag. fetched_at is UTC day granularity. Primary key: (month, package_name). Cadence: monthly. Nullability: license_id may be empty when upstream supplies none; scores and ranks are never null. Caveats: NuGet exposes cumulative download totals only, so growth rates need two snapshots and are not in this first snapshot; counts include automated and CI traffic, not just human adoption; the taxonomy is a first-match heuristic over tags and description text, not an authoritative classification; per-package license ids are author-supplied and informational only. Sample use: order by popularity_rank for the month's most-adopted data packages, or filter category = 'machine-learning'.

Les titres et les descriptions proviennent des sources de données, en anglais.