Datasets

Iris Dataset

The Iris dataset, introduced by British statistician and biologist Ronald A. Fisher in 1936, is one of the most iconic datasets in the history of machine learning and statistics. It contains measurements of 150 iris flowers, evenly split across three species — Iris setosa, Iris versicolor, and Iris virginica — with each flower described by four features: sepal length, sepal width, petal length, and petal width, all measured in centimetres. Its appeal lies in its simplicity and structure: one class (setosa) is linearly separable from the other two, while the remaining pair overlaps slightly, making it an ideal sandbox for exploring both simple and more nuanced classification algorithms.

We can load the dataset directly from the Python pandas library.

import pandas as pd
from sklearn.datasets import load_iris
import matplotlib.pyplot as plt

iris = load_iris(as_frame=True)
df = iris.frame
df["species"] = df["target"].map({0: "setosa", 1: "versicolor", 2: "virginica"})
df = df.drop(columns=["target"])

df.head(6)

	sepal length (cm)	sepal width (cm)	petal length (cm)	petal width (cm)	species
0	5.1	3.5	1.4	0.2	setosa
1	4.9	3.0	1.4	0.2	setosa
2	4.7	3.2	1.3	0.2	setosa
3	4.6	3.1	1.5	0.2	setosa
4	5.0	3.6	1.4	0.2	setosa
5	5.4	3.9	1.7	0.4	setosa

The first few rows of the Iris dataset.

Palmer Penguins Dataset

The Palmer Penguins dataset was collected by Dr. Kristen Gorman at Palmer Station, Antarctica.

Gorman KB, Williams TD, Fraser WR (2014) Ecological Sexual Dimorphism and Environmental Variability within a Community of Antarctic Penguins (Genus Pygoscelis). PLoS ONE 9(3): e90081. doi:10.1371/journal.pone.0090081

It contains measurements for 344 penguins belonging to three species:

Adélie (Pygoscelis adeliae)
Chinstrap (Pygoscelis antarcticus)
Gentoo (Pygoscelis papua)

The features available are:

Feature	Type	Description
`bill_length_mm`	Numeric	Length of the bill in millimetres
`bill_depth_mm`	Numeric	Depth (height) of the bill in millimetres
`flipper_length_mm`	Numeric	Flipper length in millimetres
`body_mass_g`	Numeric	Body mass in grams
`island`	Categorical	Biscoe, Dream, or Torgersen
`sex`	Categorical	Male or Female

We can load the dataset directly from the Python seaborn library.

# Load libraries
import seaborn as sns

# Load the dataset
penguins = sns.load_dataset("penguins")

# Show the first few rows
penguins.head()

	species	island	bill_length_mm	bill_depth_mm	flipper_length_mm	body_mass_g	sex
0	Adelie	Torgersen	39.1	18.7	181.0	3750.0	Male
1	Adelie	Torgersen	39.5	17.4	186.0	3800.0	Female
2	Adelie	Torgersen	40.3	18.0	195.0	3250.0	Female
3	Adelie	Torgersen	NaN	NaN	NaN	NaN	NaN
4	Adelie	Torgersen	36.7	19.3	193.0	3450.0	Female

Using your own dataset

With pandas, you can easily load your own datasets from various formats, such as CSV, Excel, or SQL databases.

For example, we can load the CSV file nhs_readmission.csv. This dataset is entirely synthetic, generated purely as a teaching example.

Simulated distributions based on typical NHS inpatient population parameters (e.g. age ~Normal(62, 16), length of stay ~Exponential(4.5), HbA1c ~Normal(60, 18))
Clinically plausible missingness patterns (HbA1c missing in ~16% of patients, eGFR in ~9%)
A logistic outcome model where readmission probability depends on age, Charlson index, length of stay, prior ED attendances, eGFR, and IMD decile

Column descriptions

Column	Description
`patient_id`	Anonymous patient identifier
`age`	Age in years at admission
`sex`	M / F
`imd_decile`	Index of Multiple Deprivation (1 = most deprived, 10 = least)
`primary_diagnosis`	ICD-10 code for primary diagnosis
`charlson_index`	Charlson Comorbidity Index (0–37)
`hba1c`	Most recent HbA1c (mmol/mol) — missing if not tested
`egfr`	Most recent eGFR (mL/min/1.73m²) — missing if not tested
`bmi`	Body Mass Index (kg/m²)
`los_days`	Length of stay in days
`ed_prior_year`	ED attendances in previous 12 months
`discharge_ward`	Ward the patient was discharged from
`readmitted_30d`	Readmitted within 30 days: Yes / No

import pandas as pd

url = "https://raw.githubusercontent.com/Bristol-Training/nhs-data-science/refs/heads/main/data/nhs_readmission.csv"

nhs_data = pd.read_csv(
    url,
    skiprows=3, # skip header rows with metadata
    index_col="patient_id", # this column is an identifier each row
    na_values=[".", "N/A", "NULL", "UNKNOWN"],  # recode common placeholders as NaN
)

print("Missing values per column:")
print(nhs_data.isnull().sum())

Missing values per column:
age                    0
sex                    0
imd_decile             0
primary_diagnosis      0
charlson_index         0
hba1c                811
egfr                 445
bmi                   65
los_days               0
ed_prior_year         25
discharge_ward         0
readmitted_30d         0
dtype: int64

# delete rows with missing values
nhs_data = nhs_data.dropna()

# print first rows 
nhs_data.head()

	age	sex	imd_decile	primary_diagnosis	charlson_index	hba1c	egfr	bmi	los_days	ed_prior_year	discharge_ward	readmitted_30d
patient_id
1001	69	M	8	E11	2	67.6	29.4	36.4	5	1.0	Endocrine	No
1002	59	M	2	I10	4	67.5	42.8	37.8	4	2.0	General Medicine	No
1003	72	M	2	I63	1	44.1	61.7	26.2	3	1.0	Gastro	No
1004	86	M	10	J18	1	55.5	62.9	34.0	2	1.0	General Medicine	No
1005	58	M	10	C34	2	59.8	120.0	18.5	6	0.0	Respiratory	No

Other datasets

scikit-learn example datasets https://scikit-learn.org/stable/datasets.html

Kaggle open datasets https://www.kaggle.com/datasets

Hugging face datasets https://huggingface.co/datasets