---
id: adtk
version: "0.6.2"
license: Mozilla Public License 2.0 (MPL 2.0)
license_treatment: copyleft
maintenance: dormant
---

# adtk — A package for unsupervised time series anomaly detection

License: copyleft · Maintenance: dormant · Downloads: 522.9K/mo

## What it is and what it does

ADTK is a Python library for detecting anomalies in time series data using unsupervised and rule-based methods. It provides a collection of detector algorithms, feature transformers, and ensemble aggregators that can be combined via pipe classes into complete anomaly detection models. The package also includes utilities for time series visualization and anomaly event processing.

The library is designed around the principle that no single anomaly detection algorithm works universally—instead, it offers building blocks with unified APIs so you can select and combine the right detectors, transformers, and aggregators for your specific problem. It depends on numpy, pandas, scikit-learn, statsmodels, and matplotlib for its core functionality.

Use it for:

- Detect sudden spikes or drops in sensor data or system metrics without labeled training data.
- Identify unusual patterns in financial time series using statistical rules.
- Flag anomalous behavior in network traffic or application performance logs.
- Combine multiple detection algorithms to improve sensitivity and reduce false positives in critical monitoring.
- Visualize detected anomalies alongside raw time series data for manual review and validation.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

ADTK provides detectors, transformers, and aggregators for unsupervised and rule-based time series anomaly detection, with unified APIs and pipe classes to combine them into detection models.

Yes, if you need unsupervised time series anomaly detection and can accept dormant maintenance. The package is stable (Production/Stable status), has no known vulnerabilities, and low install friction. However, the last release was 2020-04-17; evaluate whether the detectors and dependencies meet your current needs before committing to it in new projects.

## Install

pip install adtk
uv add adtk
poetry add adtk

## Installing adtk

Before you install: Low install friction with a pure-Python wheel and well-established dependencies. However, the package is dormant—last release was 2020-04-17, with no recent maintenance activity despite an active repository.

License in practice: Licensed under Mozilla Public License 2.0 (MPL 2.0), a copyleft license requiring derivative works and modifications to be distributed under the same license. Review your project's license compatibility before use.

Quickstart:

pip install adtk

from adtk.detector import ThresholdAD
import pandas as pd

detector = ThresholdAD(high=3.0)
anomalies = detector.detect(time_series)

Requires Python 3.5 or later; depends on numpy, pandas, scikit-learn, and statsmodels being installed.

Verify before relying:

- Whether the package's detectors and transformers remain effective on modern time series data given dormant maintenance.
- Compatibility with Python versions beyond 3.8, which was the latest tested version at release.

## Package facts

- License: Mozilla Public License 2.0 (MPL 2.0) (copyleft)
- Python support: supports_current
- Install friction: low
- Maintenance: dormant
- Downloads: 522.9K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

time series anomaly detection, unsupervised anomaly detection, anomaly detection toolkit, time series outlier detection, rule-based anomaly detection, anomaly detection pipeline, statistical anomaly detection, time-series, anomaly-detection, unsupervised-learning

[View on SkillFed](https://skillfed.io/packages/adtk) · [View on PyPI](https://pypi.org/project/adtk/)