changepoynt

changepoynt is a Python package for change point detection in time series. It includes subspace-based methods such as SST, IKA-SST, ESST, and MESST, density-ratio methods such as uLSIF and RuLSIF, and matrix-profile segmentation methods such as FLUSS and FLOSS.

The package focuses on efficient implementations with readable code, so it can be useful both for research experiments and practical time-series workflows.

Start Here

Install the package:

pip install changepoynt

Then run a detector on a one-dimensional time series:

import numpy as np
from changepoynt.algorithms.esst import ESST

signal = np.concatenate([
    np.ones(200),
    np.exp(-np.linspace(0, 5, 200)),
    np.exp(-5) * np.ones(150),
    0.2 * np.sin(np.linspace(0, 30 * np.pi, 300)),
])

detector = ESST(window_length=40)
score = detector.transform(signal)

See the quickstart for a complete plotted example.

Documentation Map

• Installation covers PyPI and source installs.
• Quickstart shows a minimal end-to-end example.
• Algorithms summarizes implemented methods and their status.
• User Guides explains how to choose and configure related method families.
• API Reference is generated from package docstrings.
• FAQ answers common parameter and performance questions.