Python module for change point analysis

Question

I'm looking for a Python module that performs a change-point analysis on a time-series. There are a number of different algorithms and I'd like to explore the efficacy of some of them without having to hand-roll each of the algorithms.

Ideally I'd like some modules like the bcp (Bayesian Change Point) or strucchange packages in R. I expected to find some in Scipy but I haven't been able to turn up anything.

I'm surprised that there aren't any facilities in:

• statsmodels.tsa: Time series statistical analysis tools
• scikits.timeseries: Time series analysis tools to extend scipy
• scipy.signal: signal processing tools in scipy

Are there any modules with change point detection algorithms in Python?

Answer 1

You can try out the changefinder library on PyPI. The description says that it's an online Change Detection Library based on the ChangeFinder algorithm

There are also some Python implementations of Michele Basseville's Statistical Change Point Detection techniques available in tutorial format on this Github repo.

Answer 2

There are still some gaps in the Python library for using advanced statistics packages. Have you tried using the RPy module? When using RPy you can load R modules.

brief tutorial on RPy: http://www.sciprogblog.com/2012/08/using-r-from-within-python.html strucchange

Answer 3

I just came across a change point detection library in Python named "ruptures" : https://arxiv.org/abs/1801.00826

Maybe this can be of use.

Answer 4

This implementation of the Python package rpy2 worked for me:

import numpy as np
from rpy2.robjects.packages import importr
import rpy2.robjects as robjects

r = robjects.r #allows access to r object with r.

bcp = importr('bcp') #import bayesian change point package in python

values = bcp.bcp( r.c( r.rnorm(50) , r.rnorm(50,5,1), r.rnorm(50) ) ) #use bcp function on vector

posterior_means = np.array(values[5]).flatten()
posterior_probability = np.array(values[7]).flatten()

Then, you can plot the posterior means and posterior probability against the original vector. See the bcp function example in R for more detailed information about this example.

Also, hard indexing values with a number (i.e. values[5]) is not ideal, but I was having a hard time using the rx and rx2 extractor. So if anyone can enlighten me on a less hacky method of extraction, I'd love to know!

Answer 5

Depending on your requirement for online/offline change point detection, python has the below packages:

1) The ruptures package, a Python library for performing offline change point detection.

2) Calling the R changepoint package into Python using the rpy2 package, an R-to-Python interface.

3) The changefinder package, a Python library for online change point detection.

4) Bayesian Change Point Detection - both online and offline approaches.

Answer 6

Have you tried ChangeFinder library, you can install it on linux by:

pip install changefinder

also Bayesian_changepoint_detection GitHub code can be found here: GitHub Code