What is the Wine Dataset dataset?

Classic UCI Wine Dataset with 178 samples from 3 Italian cultivars, featuring 13 chemical attributes for multi-class classification, clustering, and dimensionality reduction benchmarking.

How can I download the Wine Dataset dataset?

You can download the Wine Dataset dataset in CSV or Parquet format directly from this page. Each table has its own download buttons.

What format is the Wine Dataset dataset available in?

This dataset is available in CSV and Parquet formats. CSV is great for spreadsheet applications, while Parquet is optimized for data analysis tools like Pandas and DuckDB.

What is the license for the Wine Dataset dataset?

This dataset is available under the CC BY 4.0 license. See the full license at https://creativecommons.org/licenses/by/4.0/

How many tables are in the Wine Dataset dataset?

The Wine Dataset dataset contains 1 table: wine.

Wine Dataset

Last updated 1 weeks ago•January 2, 2026

Source:UCI Machine Learning Repository

License:CC BY 4.0

Version:1.0

Time:1980s

Location:Piedmont, Italy

Created by Dataset Agent

Overview

The Wine Dataset is one of the most iconic benchmark datasets in machine learning, originating from chemical analysis of wines grown in the Piedmont region of Italy. The wines derive from three different cultivars (grape varieties), and the classification task involves determining which cultivar a wine belongs to based solely on its chemical properties. Created by M. Forina and colleagues at the Institute of Pharmaceutical and Food Analysis and Technologies in Genoa, Italy, this dataset was donated to the UCI Machine Learning Repository in 1991 by Stefan Aeberhard.

The dataset contains 178 wine samples with complete chemical analysis profiles across 13 continuous attributes and 3 cultivar classes. <details> <summary>View SQL Query</summary> ``

sql
SELECT COUNT(*) as total_samples, COUNT(DISTINCT class) as num_classes FROM wine.csv

`` </details>

View Source

SQL

SELECT COUNT(*) AS row_count FROM wine.csv

Data

Row Count
178
1 row

Zero Missing Values: Unlike many real-world datasets, the Wine Dataset is complete with no missing values across all 178 samples and 14 columns. This makes it ideal for learning ML fundamentals without requiring imputation techniques.

The 13 Chemical Attributes Explained

Understanding what each chemical attribute measures is crucial for interpreting model results and feature importance. These attributes capture the essential chemical fingerprint that distinguishes wines from different cultivars.

Class Distribution Analysis

The three wine cultivars show moderate class imbalance, which is important to consider when training classifiers and evaluating model performance.

View Source

SQL

SELECT class, COUNT(*) AS count FROM wine.csv GROUP BY class ORDER BY class

Data

Cultivar Class	Sample Count	Percentage
Class 1	59	33.1%
Class 2	71	39.9%
Class 3	48	27.0%
3 rows

Class 2 dominates with 71 samples (39.9%), while Class 3 has the fewest with 48 samples (27.0%). This ~1.5:1 imbalance ratio is mild but should be considered when using accuracy as an evaluation metric. <details> <summary>View SQL Query</summary> ``

sql
SELECT class, COUNT(*) as count, ROUND(COUNT(*) * 100.0 / 178, 1) as percentage FROM wine.csv GROUP BY class ORDER BY class

`` </details>

View Source

SQL

SELECT class, COUNT(*) AS count FROM wine.csv GROUP BY class ORDER BY class

Data

Class	Count
1	59
2	71
3	48
3 rows

Feature Statistics Summary

The 13 features span vastly different scales, making feature scaling (standardization or normalization) essential for distance-based algorithms like k-NN and SVM.

Descriptive Statistics for All Features

Feature	Min	Max	Mean	Std Dev	Scale Factor
Alcohol	11.03	14.83	13.00	0.81	1x
Malic Acid	0.74	5.80	2.34	1.12	1x
Ash	1.36	3.23	2.37	0.27	1x
Alcalinity of Ash	10.60	30.00	19.49	3.34	10x
Magnesium	70	162	99.74	14.28	100x
Total Phenols	0.98	3.88	2.29	0.63	1x
Flavanoids	0.34	5.08	2.03	1.00	1x
Nonflavanoid Phenols	0.13	0.66	0.36	0.12	0.1x
Proanthocyanins	0.41	3.58	1.59	0.57	1x
Color Intensity	1.28	13.00	5.06	2.32	10x
Hue	0.48	1.71	0.96	0.23	1x
OD280/OD315	1.27	4.00	2.61	0.71	1x
Proline	278	1680	746.89	314.91	1000x
13 rows

View Source

SQL

SELECT ROUND(MIN(alcohol), 2), ROUND(MAX(alcohol), 2), ROUND(AVG(alcohol), 2), ROUND(STDDEV (alcohol), 2) FROM wine.csv

Data

Feature	Min	Max	Mean	Std Dev	Scale Factor
Alcohol	11.03	14.83	13.00	0.81	1x
Malic Acid	0.74	5.80	2.34	1.12	1x
Ash	1.36	3.23	2.37	0.27	1x
Alcalinity of Ash	10.60	30.00	19.49	3.34	10x
Magnesium	70	162	99.74	14.28	100x
Total Phenols	0.98	3.88	2.29	0.63	1x
Flavanoids	0.34	5.08	2.03	1.00	1x
Nonflavanoid Phenols	0.13	0.66	0.36	0.12	0.1x
Proanthocyanins	0.41	3.58	1.59	0.57	1x
Color Intensity	1.28	13.00	5.06	2.32	10x
Hue	0.48	1.71	0.96	0.23	1x
OD280/OD315	1.27	4.00	2.61	0.71	1x
Proline	278	1680	746.89	314.91	1000x
13 rows

Proline shows the largest variation with values ranging from 278 to 1,680 mg/L (6x range), while Ash is the most consistent feature with only a 1.87 range (1.36 to 3.23). <details> <summary>View SQL Query</summary> ``

sql
SELECT 
  ROUND(MIN(proline), 0) as proline_min, 
  ROUND(MAX(proline), 0) as proline_max,
  ROUND(MIN(ash), 2) as ash_min,
  ROUND(MAX(ash), 2) as ash_max
FROM wine.csv

`` </details>

View Source

SQL

SELECT MIN(proline), MAX(proline), MIN(ash), MAX(ash) FROM wine.csv

Data

Min Proline	Max Proline	Min Ash	Max Ash
278	1,680	1.36	3.23
1 row

Distinguishing Chemical Signatures by Cultivar

Each cultivar exhibits distinct chemical profiles that enable high classification accuracy. These differences reflect unique terroir characteristics and grape variety genetics.

Average Feature Values by Cultivar Class

Feature	Class 1	Class 2	Class 3	Best Discriminator?
Alcohol (%)	13.74	12.28	13.15	✓ Class 2 lowest
Flavanoids	2.98	2.08	0.78	✓✓ Strong separator
Color Intensity	5.53	3.09	7.40	✓✓ Strong separator
Proline (mg/L)	1116	520	630	✓✓ Class 1 highest
Hue	1.06	1.06	0.68	✓ Class 3 lowest
OD280/OD315	3.16	2.78	1.68	✓ Class 3 lowest
6 rows

View Source

SQL

SELECT class, ROUND(AVG(alcohol), 2), ROUND(AVG(flavanoids), 2), ROUND(AVG(color_intensity), 2), ROUND(AVG(proline), 0) FROM wine.csv GROUP BY class

Data

Feature	Class 1	Class 2	Class 3	Best Discriminator?
Alcohol (%)	13.74	12.28	13.15	✓ Class 2 lowest
Flavanoids	2.98	2.08	0.78	✓✓ Strong separator
Color Intensity	5.53	3.09	7.40	✓✓ Strong separator
Proline (mg/L)	1116	520	630	✓✓ Class 1 highest
Hue	1.06	1.06	0.68	✓ Class 3 lowest
OD280/OD315	3.16	2.78	1.68	✓ Class 3 lowest
6 rows

Flavanoids is the strongest discriminating feature: Class 1 averages 2.98, Class 2 averages 2.08, and Class 3 averages only 0.78 — nearly a 4x difference between Class 1 and Class 3. <details> <summary>View SQL Query</summary> ``

sql
SELECT class, ROUND(AVG(flavanoids), 2) as avg_flavanoids FROM wine.csv GROUP BY class ORDER BY class

`` </details>

View Source

SQL

SELECT class, ROUND(AVG(flavanoids), 2) AS avg_flavanoids FROM wine.csv GROUP BY class ORDER BY class

Data

Class	Avg Flavanoids
1	2.98
2	2.08
3	0.78
3 rows

View Source

SQL

SELECT class, ROUND(AVG(flavanoids), 2) AS avg_flavanoids FROM wine.csv GROUP BY class ORDER BY class

Data

Cultivar	Avg Flavanoids
Class 1	2.98
Class 2	2.08
Class 3	0.78
3 rows

Feature Correlations and Multicollinearity

Several features are highly correlated, which affects feature selection strategies and can cause issues with certain algorithms. Understanding these relationships guides effective dimensionality reduction.

Flavanoids and Total Phenols show the strongest positive correlation at r = 0.865, indicating these phenolic compounds are chemically related. This high correlation suggests one could be dropped without significant information loss. <details> <summary>View SQL Query</summary> ``

sql
SELECT ROUND(CORR(flavanoids, total_phenols), 3) as correlation FROM wine.csv

`` </details>

View Source

SQL

SELECT ROUND(CORR(flavanoids, total_phenols), 3) AS flavanoids_phenols_corr FROM wine.csv

Data

Flavanoids Phenols Corr
0.87
1 row

Notable Feature Correlations

Feature Pair	Correlation	Interpretation
Flavanoids ↔ Total Phenols	+0.865	Strong positive; redundant information
Flavanoids ↔ OD280/OD315	+0.787	Both relate to phenolic content
Alcohol ↔ Proline	+0.644	Moderate positive; ripeness indicators
Hue ↔ Malic Acid	+0.561	Color relates to acid profile
Color Intensity ↔ Flavanoids	-0.172	Weak negative; interesting inverse
5 rows

View Source

SQL

SELECT ROUND(CORR(alcohol, proline), 3), ROUND(CORR(flavanoids, total_phenols), 3) FROM wine.csv

Data

Feature Pair	Correlation	Interpretation
Flavanoids ↔ Total Phenols	+0.865	Strong positive; redundant information
Flavanoids ↔ OD280/OD315	+0.787	Both relate to phenolic content
Alcohol ↔ Proline	+0.644	Moderate positive; ripeness indicators
Hue ↔ Malic Acid	+0.561	Color relates to acid profile
Color Intensity ↔ Flavanoids	-0.172	Weak negative; interesting inverse
5 rows

Sample Data Preview

Here are representative samples from each cultivar class showing actual data values. Note how the chemical profiles differ systematically between classes.

Sample Wine Records (Selected Features)

Class	Alcohol	Malic Acid	Flavanoids	Color Intensity	Proline
1	14.23	1.71	3.06	5.64	1,065
1	13.2	1.78	2.76	4.38	1,050
2	12.37	0.94	2.76	3	520
2	12.33	1.1	2.43	2.2	680
3	12.86	1.35	0.64	7.65	720
3	13.11	1.01	0.75	10.8	630
6 rows

View Source

SQL

SELECT class, alcohol, malic_acid, flavanoids, color_intensity, proline FROM wine.csv WHERE class = 1 LIMIT 2 UNION ALL SELECT class, alcohol, malic_acid, flavanoids, color_intensity, proline FROM wine.csv WHERE class = 2 LIMIT 2 UNION ALL SELECT class, alcohol, malic_acid, flavanoids, color_intensity, proline FROM wine.csv WHERE class = 3 LIMIT 2

Data

Class	Alcohol	Malic Acid	Flavanoids	Color Intensity	Proline
1	14.23	1.71	3.06	5.64	1,065
1	13.2	1.78	2.76	4.38	1,050
2	12.37	0.94	2.76	3	520
2	12.33	1.1	2.43	2.2	680
3	12.86	1.35	0.64	7.65	720
3	13.11	1.01	0.75	10.8	630
6 rows

Common Confusion: Wine vs. Wine Quality Dataset

Don't confuse these two popular datasets! The UCI Wine Dataset (this one) and the Wine Quality Dataset are frequently mixed up but serve different purposes.

Expected Classification Performance

Due to well-separated class boundaries, this dataset achieves high accuracy with most classifiers. Here are typical benchmark results:

The high achievable accuracy (95-100%) makes this dataset excellent for validating implementations and learning ML workflows, but less suitable for benchmarking state-of-the-art algorithms where performance differences would be negligible.

Historical Context and Provenance

The Wine Dataset originates from research conducted in the 1980s at the Institute of Pharmaceutical and Food Analysis and Technologies in Genoa, Italy. The wines analyzed came from the Piedmont region, one of Italy's most prestigious wine-producing areas, known for varieties like Barolo, Barbaresco, and Barbera. While the specific cultivar names are not disclosed in the dataset, they represent three distinct grape varieties grown in the same geographical region.

Stefan Aeberhard donated the dataset to the UCI Machine Learning Repository in 1991, where it has since been cited in thousands of academic papers. The original research was published in Aeberhard, S., Coomans, D., and de Vel, O. (1994) "Comparative analysis of statistical pattern recognition methods in high dimensional settings" in Pattern Recognition, 27(8):1065-1077.

Limitations and Considerations

Small Sample Size: With only 178 samples, results may not generalize to larger wine classification challenges
Regional Specificity: All wines come from Piedmont, Italy — models may not transfer to wines from other regions
Historical Data: Chemical analysis methods from the 1980s may differ from modern analytical techniques
Class Imbalance: Class 3 has 32% fewer samples than Class 2 (48 vs. 71), which may bias some classifiers
Unknown Cultivars: The specific grape varieties are not disclosed, limiting domain interpretation
"Too Easy" for Benchmarking: Near-perfect accuracy is achievable, making it unsuitable for comparing advanced algorithms

Table Overview

wine

Contains 178 rows and 14 columns. Column types: 14 numeric.

178 rows14 columns

wine

178

rows

columns

Data Preview

Scroll to see more

class	alcohol	malic_acid	ash	alcalinity_of_ash	magnesium	total_phenols	flavanoids	nonflavanoid_phenols	proanthocyanins	color_intensity	hue	od280_od315	proline
1	14.23	1.71	2.43	15.6	127	2.8	3.06	0.28	2.29	5.64	1.04	3.92	1,065
1	13.2	1.78	2.14	11.2	100	2.65	2.76	0.26	1.28	4.38	1.05	3.4	1,050
1	13.16	2.36	2.67	18.6	101	2.8	3.24	0.3	2.81	5.68	1.03	3.17	1,185
1	14.37	1.95	2.5	16.8	113	3.85	3.49	0.24	2.18	7.8	0.86	3.45	1,480
1	13.24	2.59	2.87	21	118	2.8	2.69	0.39	1.82	4.32	1.04	2.93	735

Row 1

class1

alcohol14.23

malic_acid1.71

ash2.43

alcalinity_of_ash15.6

+9 more columns

Row 2

class1

alcohol13.2

malic_acid1.78

ash2.14

alcalinity_of_ash11.2

+9 more columns

Row 3

class1

alcohol13.16

malic_acid2.36

ash2.67

alcalinity_of_ash18.6

+9 more columns

Showing 5 of 178 rows

Data Profile

178

rows

columns

100%

complete

121.7 KB

estimated size

Column Types

14 Numeric

High-Cardinality Columns

Columns with many unique values (suitable for identifiers or categorical features)

malic_acid(133 unique values)
flavanoids(132 unique values)
color_intensity(132 unique values)
alcohol(126 unique values)
od280_od315(122 unique values)
proline(121 unique values)
proanthocyanins(101 unique values)
total_phenols(97 unique values)

Data Dictionary

wine

Column	Type	Description	Example
`class`	numeric	Class category	1, 1
`alcohol`	numeric	Numeric value (range: 13.16 - 14.37)	14.23, 13.2
`malic_acid`	numeric	Unique identifier	1.71, 1.78
`ash`	numeric	Numeric value (range: 2.14 - 2.87)	2.43, 2.14
`alcalinity_of_ash`	numeric	Numeric value (range: 11.2 - 21)	15.6, 11.2
`magnesium`	numeric	Numeric value (range: 100 - 127)	127, 100
`total_phenols`	numeric	Numeric value (range: 2.65 - 3.85)	2.8, 2.65
`flavanoids`	numeric	Numeric value (range: 2.69 - 3.49)	3.06, 2.76
`nonflavanoid_phenols`	numeric	Numeric value (range: 0.24 - 0.39)	0.28, 0.26
`proanthocyanins`	numeric	Numeric value (range: 1.28 - 2.81)	2.29, 1.28
`color_intensity`	numeric	Numeric value (range: 4.32 - 7.8)	5.64, 4.38
`hue`	numeric	Numeric value (range: 0.86 - 1.05)	1.04, 1.05
`od280_od315`	numeric	Numeric value (range: 2.93 - 3.92)	3.92, 3.4
`proline`	numeric	Numeric value (range: 735 - 1,480)	1065, 1050