Hugging Face's logo

Datasets:
dragonlimited
/
Dragon-Finance-RealEstate

Languages:
English
Tags:
finance
financial-data
us-financial-data
realestate
License:
Dataset card Files
xet
 Community
Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 84, in _split_generators
                  raise ValueError(
                  ...<2 lines>...
                  )
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Dragon Finance — realestate

stored: 382.13 GB · shards: 234 · class: realestate · sources: US + global official

US housing and property — prices, construction, rents and housing microdata.

Part of the Dragon Finance Corpus — public financial data from US federal agencies, world official institutions and public registries. Nothing licensed, paywalled or registration-only.

Front door: Dragon-Finance-Corpus — every card, the chunked corpus and the derived artefacts, all measured from the hub; collection — this repo and its siblings.

How this was built, and what comes next: ROADMAP.md - the pipeline's phases, its operating procedure, and the sources it will never take.

Coverage

Prices FHFA house price index, Zillow research indices (commercial, non-official)
Construction Census Building Permits Survey, 1959–present
Rents and affordability HUD fair market rents
Microdata Census American Housing Survey public-use files

Size

metric value
stored 382.13 GB
shards 234
training-ready tokens none yet - no text from this class is in the chunked corpus yet, so it has no training-ready tokens to count

Training-ready text

Measured 2026-11-06 19:58 UTC over 84 of 84 shard(s). Tokens use the one convention these cards share - chars / 4 over the chunked text - so each figure is reconcilable against the shards.

source chunks text tokens share
(none yet - no text from this class is in the chunked corpus)

Class attribution uses the shard path's first segment, so a tier that shares a segment with another class (census, federalreserve, sec) is attributed to its dominant class.

Sources present

source jurisdiction data licence
US Census US trade, housing, business dynamics public
openFEMA US NFIP policies and claims public
FHFA US house price index public
FRED (St. Louis Fed) US rates + macro series CC-BY-4.0
progress.png n/a n/a n/a
Zillow research US home value / rent indices commercial - NON-OFFICIAL

What you can do with it

  • Model house-price dynamics
  • Track construction and permitting cycles
  • Assess rental affordability by area

How this class was gathered

  • FHFA house price index; US Census construction, building permits (back to 1959) and housing surveys.
  • ACS housing tables and ACS PUMS microdata - person and household records, about 125 GB across 29 years, which is the largest single source in this class.
  • Zillow research indices - commercial and NON-OFFICIAL, and labelled as such; they are here for coverage that no official source publishes at ZIP level.
  • HUD fair market rents are not included: the portal answers a bot-protection response to every keyless client, so the tier reports it as unreachable rather than pretending to hold it.

Source selection, source terms, sharding, the record that makes the corpus resumable, and what is excluded are documented once on the front door.

How to analyse it

  • PUMS CSVs have no header in older vintages - the column layout is published separately as a file list, so map positions to names yourself; newer vintages differ from older ones.
  • ACS summary tables are key-value JSON per geography; Zillow files are wide (one row per region, one column per month), so transpose before time-series work.
  • Join on FIPS (state/county/ZIP) or CBSA; Zillow uses its own region ids.

For engineers

Data is stored as tar shards (the hub caps a directory at 10,000 files and these tiers run to millions). Members keep their original names and relative paths.

import tarfile
with tarfile.open("0000.tar") as tf:
    print(tf.getnames()[:5])
    tf.extractall("out/")
shard size
README.md 8.2 KB
census/american-housing-survey/0000.tar 883.14 MB
census/building-permits-state/0000.tar 655.4 KB
census/building-permits-state/0001.tar 675.8 KB
census/building-permits-state/0002.tar 665.6 KB
census/building-permits-state/0003.tar 645.1 KB

Showing 6 of 234 shards.

Every shard records its sha256 in the publication record; shards are append-only, so a resumed run re-uploads only what changed. A shard with a failed member is never marked complete, so missed files are retried rather than dropped.

Not advice, no warranty, no endorsement. Compiled from third-party public sources and provided 'as is': derived and model-generated content may be wrong, and source material is routinely restated, amended or superseded. No source agency, exchange or company has reviewed, approved or endorsed this dataset, and none is affiliated with its publisher. Do not use it to re-identify any person, to manipulate any market, or as the sole basis for any decision about any person, entity or instrument. Terms of use: DISCLAIMER.md.
Downloads last month
829
Total file size:
382 GB

Collection including dragonlimited/Dragon-Finance-RealEstate