Classifier prototyping
Quickly obtain a small, well-known labeled dataset for testing a classifier.
Scikit-Learn · Function
load_iris
Loads scikit-learn’s built-in Iris dataset for multi-class classification. It returns a metadata-rich Bunch by default, or data and targets directly with configurable pandas output.
- Since
- 0.18
How it works
load_iris provides the built-in Iris classification dataset. By default it returns a dictionary-like Bunch containing feature data, labels, names, and a description; optional flags can instead return feature and target objects directly and use pandas containers.
Syntax
sklearn.datasets.load_iris(*, return_X_y=False, as_frame=False)Parameters
| Name | Type | Description |
|---|---|---|
return_X_y | bool, default=False | Whether to return a tuple of data and target instead of a Bunch object. |
as_frame | bool, default=False | Whether to return pandas DataFrame or Series objects with appropriate dtypes. |
Returns
By default, a Bunch dictionary-like object containing the dataset and metadata. With return_X_y=True, returns (data, target); with as_frame=True, the data and target use pandas DataFrame or Series representations as documented.
Key points
- The function supplies the built-in Iris dataset for a classification task.
- Iris is presented as a classic, easy multi-class classification dataset.
- The feature component is two-dimensional, with rows representing samples and columns representing features.
- Rows correspond to individual observations, while columns correspond to measured features.
- return_X_y=True provides data and target directly as a two-item tuple rather than the metadata-rich Bunch.
- as_frame=True makes the feature data a pandas DataFrame with numeric column dtypes.
Common mistakes
- Setting return_X_y=True changes the return type from one Bunch to a two-item tuple, so code must unpack the result.
- With as_frame=True, the target's pandas type depends on its number of columns; do not assume it is always a DataFrame.
Example: Load Iris as a DataFrame
from sklearn.datasets import load_iris
iris = load_iris(as_frame=True)
print(iris.data.shape)
print(iris.target_names.tolist())Loads the Iris dataset with feature data in a pandas DataFrame and prints its dimensions and class names.
When to use it
Tabular exploration
Use named numeric columns and pandas operations during exploratory analysis.
Teaching and demonstrations
Demonstrate dimensionality reduction or visualization on a standard dataset.
Interview questions
QWhat does load_iris return by default?
The function loads the Iris dataset for classification and, by default, returns a Bunch. The Bunch exposes feature data, classification targets, names, and dataset documentation.
Follow-up: How would you obtain pandas objects? Call load_iris(as_frame=True); the feature data becomes a pandas DataFrame and the target is a pandas DataFrame or Series as documented.
QWhat is the purpose of return_X_y?
return_X_y=False is the default and returns the complete Bunch. Setting it to True returns a tuple containing data followed by target, which is convenient when metadata is unnecessary.
Follow-up: What happens if both return_X_y=True and as_frame=True are used? The tuple elements are pandas DataFrames or Series as described by the documentation.
QWhat does as_frame change?
as_frame=True requests pandas representations with appropriate numeric column dtypes. This is useful when downstream code expects DataFrame-style labels and tabular operations.
Follow-up: Does as_frame=True alone remove the Bunch? No; return_X_y controls whether the result is a tuple, while as_frame controls the representations of its data and target components.
Related
sklearn.datasets.load_winesklearn.datasets.load_digitsThe Iris DatasetPrincipal Component Analysis (PCA) on Iris Dataset