dsa-datasets / README.md
lschiemanowski's picture
Separate technical data guide from concise dataset card
d632287 verified
|
Raw History Blame Contribute Delete
2.95 kB
metadata
license: cc-by-4.0
pretty_name: dsa datasets

dsa demo datasets

Three DuckDB databases and accompanying quantitative problem sets for dsa, an agent that answers data-analysis questions and can provide executable Jupyter notebooks explaining its answers.

Each database comes with 100 evaluation problems, including reference answers and answer schemas. The databases and problems can also be used independently of dsa.

Databases

Online Retail II

Transactions from a UK-based online retailer between December 2009 and December 2011. Around one million invoice lines contain products, quantities, prices, timestamps, customer identifiers, and countries, supporting questions about sales, purchasing patterns, and customer retention.

Source: Chen, D. (2012), Online Retail II, UCI Machine Learning Repository. Source data licensed under CC BY 4.0.

All problems · Attribution

SMARD DE/LU 2024

Electricity-market data for the Germany–Luxembourg market area in 2024. The database combines generation by energy source, electricity demand, forecasts, day-ahead prices, and cross-border trade on a 15-minute timeline, with daily summaries.

Source: Bundesnetzagentur | SMARD.de. Source data licensed under CC BY 4.0.

All problems · Attribution

EEA Air Quality — Six Cities, 2018–2024

Air-quality observations for Berlin, Madrid, Paris, Rome, Stockholm, and Warsaw. Approximately 18.6 million measurement rows cover six pollutants: SO₂, PM₁₀, O₃, NO₂, CO, and PM₂.₅. The database includes sampling-point measurements and hourly, daily, monthly, and annual city aggregates; pollutant coverage varies by city.

Source: European Environment Agency (EEA), Air Quality Download Service. Source data licensed under CC BY 4.0. The EEA does not endorse this derived dataset.

All problems · Attribution

The dsa-authored problem sets and accompanying documentation are also distributed under CC BY 4.0. Each database folder includes detailed source attribution.

Example problems

  • Online Retail II: What proportion of each acquisition cohort is still buying after one, three, six, and twelve months?
  • SMARD: On which day was renewable generation largest relative to grid load?

The linked problem lists include the calculation details and required answer fields for each question.

See the data guide for the folder layout, JSON formats, database conventions, and download instructions.