In the world of data analysis, detecting changes in data streams is a crucial task. This article will guide you through performing online (or real-time) changepoint detection using Python’s changepoint_online package, to identify abrupt shifts in your data streams as quickly as they arrive.

Changepoint detection is a statistical technique used to pinpoint moments in a time series where the underlying characteristics of the data significantly change, such as a change in the mean, variance, or distribution of the data. This can be invaluable in various applications, from monitoring financial markets for sudden price fluctuations to tracking sensor readings for equipment failures.

We just released a new software suite in Python, implementing a range of changepoint_algorithms!

changepoint_online contains all Focus exponential family algorithms as well as the NPFocus algorithm for non-parametric changepoint detection.

It’s versatile enough to be applied in scenarios where the pre-change parameter is either known or unknown.

It is possible to apply constraints to detect specific types of changes (such as increases or decreases in parameter values).

Check it out here on the github page or install via:

We uploaded on arxiv a new preprint, where we establish, through a relaxation argument, a connection between computationnal geometry and changepoint detection via functional pruning (yes, it’s another focus extension). Via this link we show that it’s possible to extend functional pruning to multidimensional cases (exactly), and we provide a comprehensive description of the conditions required for functional pruning to be applicable :)

We have recently pushed two new FOCuS extensions to arXiv.

• NP-FOCuS: This is a non-parametric version of FOCuS used to detect a change-in-distribution. The algorithm combines previous ideas that we explored in non-parametric changepoint detection with functional pruning to improve both detection power and computational complexity. It works for a variety of non-standard scenarios and is perfect for cases where the nature is unknown a priori! You can find the preprint here.

Our latest online nonparametric changepoint detection algorithm just appeared in CSDA. An R package implementing the method is available on Github at the following link.