labelfree 1.0.0

Label-free metrics for unsupervised anomaly detector evaluation

labelfree

Label-free metrics for comparing unsupervised anomaly detectors and hyperparameters when labeled anomalies are unavailable.

Install

pip install labelfree

Quick start

from labelfree.metrics import score_cluster_metrics

scores = [0.1, 0.2, 0.3, 3.8, 4.1]
metrics = score_cluster_metrics(scores, n_outliers=2)
print(metrics["silhouette"])  # higher is better

The package includes score-cluster, ASI/ASOI, AutoUAD, Laplacian, consensus, ranking-stability, IREOS, SIREOS, and Excess-Mass/Mass-Volume metrics. Every raw-score API requires an explicit score polarity when larger values mean more normal.

Label-free metrics are model-selection signals, not substitutes for labeled evaluation. Check each metric's assumptions and direction before comparing results.

See the documentation for metric selection, API details, and examples.

Development

python -m pip install -e ".[dev]"
pytest

Contributions are welcome; see CONTRIBUTING.md. Released under the MIT License.