Datasets

The library includes 22 pre-configured datasets for counterfactual explanation research, covering both classification and regression tasks.

Dataset Categories

Classification Datasets

15 datasets for binary and multi-class classification tasks.

View Datasets →

Regression Datasets

7 datasets for continuous prediction tasks.

View Datasets →

Custom Datasets

Add your own datasets with YAML configuration.

Learn How →

Quick Start

from counterfactuals.datasets import FileDataset

# Load a pre-configured dataset
dataset = FileDataset(config_path="config/datasets/adult.yaml")

# Access data
X_train, X_test = dataset.X_train, dataset.X_test
y_train, y_test = dataset.y_train, dataset.y_test

# Get feature information
print(f"Features: {dataset.features}")
print(f"Numerical: {dataset.numerical_features}")
print(f"Categorical: {dataset.categorical_features}")
print(f"Actionable: {dataset.actionable_features}")

Dataset Features

All datasets support:

Feature	Description
Automatic splitting	80/20 train/test split (stratified for classification)
Feature typing	Numerical and categorical feature distinction
Actionability	Mark which features can be modified
Constraints	Define bounds and monotonicity constraints
Cross-validation	Built-in CV split generation

Available Datasets

Classification

Dataset	Features	Classes	Size	Domain
Adult	14	2	48K	Income
COMPAS	12	2	7K	Recidivism
German Credit	20	2	1K	Credit
HELOC	23	2	10K	Credit
Give Me Some Credit	10	2	150K	Credit
...

Regression

Dataset	Features	Size	Domain
Concrete	8	1K	Engineering
Wine Quality	11	6K	Food
Diabetes	10	442	Health
...

See the full list in Classification Datasets and Regression Datasets.