Wine provenance
The exact UCI Wine recognition source, class mapping, metadata limits, and attribution.
Attribution: Aeberhard, S. & Forina, M. (1992). Wine [Dataset]. UCI Machine Learning Repository. doi:10.24432/C5PC7J.
License: CC BY 4.0, as stated on the official UCI page. The complete license text is included. The original wine.names retains credits to Forina and colleagues/PARVUS, Stefan Aeberhard, and Riccardo Leardi for attribute information.
Downloaded from the official Wine archive on October 10, 2026. The archive and extracted files are unchanged. This is Wine recognition, predicting three cultivar classes from chemical measurements; it is not the separate Wine Quality dataset.
| Property | Value |
|---|---|
| Version | uci-wine.data-sha256-6be6b1203f3d |
| Selected file | wine.data |
| File size | 10,782 bytes |
SHA-256 | 6be6b1203f3d51df0b553a70e57b8a723cd405683958204f96d23d7cd6aea659 |
| Observations / features | 178 / 13 |
| Header / delimiter | No header / comma |
| Missing or nonfinite feature fields | None in the pinned file |
| Distinct complete record lines | 178 |
| Task | Three class cultivar classification |
Feature and label order
The first CSV field is the class code. The following thirteen fields are features:
| Zero based feature index | Aner name | Measurement | Unit metadata |
|---|---|---|---|
| 0 | alcohol | Alcohol | Not supplied |
| 1 | malic_acid | Malic acid | Not supplied |
| 2 | ash | Ash | Not supplied |
| 3 | alcalinity_of_ash | Alcalinity of ash | Not supplied |
| 4 | magnesium | Magnesium | Not supplied |
| 5 | total_phenols | Total phenols | Not supplied |
| 6 | flavanoids | Flavanoids | Not supplied |
| 7 | nonflavanoid_phenols | Nonflavanoid phenols | Not supplied |
| 8 | proanthocyanins | Proanthocyanins | Not supplied |
| 9 | color_intensity | Color intensity | Not supplied |
| 10 | hue | Hue | Not supplied |
| 11 | od280_od315_of_diluted_wines | OD280/OD315 of diluted wines | Not supplied |
| 12 | proline | Proline | Not supplied |
Official supplied files and current variable metadata do not specify measurement units. Aner returns unknown; the manifest stores null with an unavailable status. These values do not mean dimensionless units. All runtime features use Float64, including integer valued measurements such as magnesium.
| Source code | Aner class index | Display name | Rows |
|---|---|---|---|
1 | 0 | class_1 | 59 |
2 | 1 | class_2 | 71 |
3 | 2 | class_3 | 48 |
Cultivar names are not provided: these display names are identifiers. Original row order is retained and grouped by class. Some source decimals omit a leading zero, such as .28; the data reader handles these separately from Aner's source literal syntax.
Changes: downloaded files are unchanged. Aner parses features as Float64, maps class codes to zero based indices, and assigns source row IDs. It does not impute, standardize, drop, or reorder observations by default.
This small historical sample covers one Italian region and three cultivars. No official train/test split is provided. The task does not predict quality scores. Unequal feature scales make explicit, preprocessing fitted on training data useful for some models. The manifest records archive and metadata hashes.