Iris dataset: Fisher, R. (1936). Iris [Dataset]. UCI Machine Learning Repository.
DOI: https://doi.org/10.24432/C56C76
Source: https://archive.ics.uci.edu/dataset/53/iris
Retrieved 2026-10-11. Distributed under CC BY 4.0:
https://creativecommons.org/licenses/by/4.0/

The archived bezdekIris.data is unchanged. iris.csv adds only column names and
generated row IDs; its four measurements retain centimetre units. UCI documents
historical differences between iris.data and Fisher's paper; this lab explicitly
uses the supplied bezdekIris.data variant. It does not silently correct a source.

Teaching alterations are made in memory and are NOT defects claimed to exist
in the original data: each row has two copies; every tenth source row is missing
petal length; every thirteenth source row has sepal length expressed in mm.
The two copies share their source row and defects. Recorded units permit exact
conversion; no heuristic guesses units. Grouping identical original measurements
is a conservative teaching rule, not evidence of biological specimen identity.

Provided as-is, without warranties; no endorsement by the creator or UCI.
See manifest.json for hashes, preprocessing and split policy.