Example Datasets

pgmpy ships a curated collection of benchmark datasets for causal discovery and graphical modeling. Each dataset includes not only the data table, but also metadata such as ground-truth graphs, expert knowledge, and dataset characteristics — ready to use without any manual download or preparation.

Tip

When to use this vs. Example Models: Example Datasets provide data tables for learning graph structures and parameters. Example Models provide pre-built graph structures (and optionally parameters) for inference, simulation, and benchmarking.

At a Glance

• Unified API: Discover datasets with list_datasets(...) and load them with load_dataset(...).

• Rich Filtering: Filter datasets by type, size, and available metadata before loading.

• Bundled Metadata: Each dataset carries its ground-truth graph, expert knowledge, and tags alongside the data.

API

The dataset API has two entry points — discover, then load:

from pgmpy.datasets import list_datasets, load_dataset

matches = list_datasets(is_discrete=True, has_ground_truth=True)
dataset = load_dataset(matches[0])

print(dataset.name)
print(dataset.data.shape)
print(dataset.tags)

The loaded object exposes the data as a pandas.DataFrame and keeps all metadata on the same object for downstream workflows.

Filtering

list_datasets(**filters) narrows the catalog before loading. Supported filters include data type (is_discrete, is_continuous, is_mixed), available metadata (has_ground_truth, has_expert_knowledge), size (n_variables, n_samples), and data origin (is_simulated, is_interventional).

Bundled Metadata

Each loaded dataset provides structured access to:

• data: The tabular data as a pandas.DataFrame, ready for estimator.fit(data).

• ground_truth: The true causal graph (when available), for use with supervised metrics.

• expert_knowledge: Domain constraints (when available), for use with discovery algorithms.

• tags: Dictionary of dataset properties and characteristics.

See Also

See also

• Causal Discovery — Use loaded datasets to learn causal graphs.

• Example Models — Pre-built graph structures to pair with these datasets.

API Reference

For the full list of available datasets:

• Datasets and Example Models API