Metadata-Version: 2.1
Name: ChangePointDetectorEVT
Version: 0.0.1
Summary: This module takes a time series and returns: (a) the underlaying linear trend and (b) the times where there is a change in the trend
Home-page: https://github.com/mhaupt63/ChangePointDetector
Author: Michael Hauptman
Author-email: michael.hauptman1@gmail.com
License: UNKNOWN
Project-URL: Bug Tracker, https://github.com/mhaupt63/ChangePointDetectorissues
Keywords: ['Time Series','Anomaly','Detection','Kalman','EVT']
Platform: UNKNOWN
Classifier: Development Status :: 3 - Alpha
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.7
Description-Content-Type: text/markdown
License-File: LICENSE.txt

ChangePointDetector
This module takes a time series and returns: (a) the piecewise underlaying linear trend, and (b) the times where there is a change in the underlying trend

We use a Kalman Filter to find the piecewise underlying linear trend, with a state space representation of seasonality and an underlying linear trend. We initialise the parameters of the Kalman filter using a Least Squares estimator out of "statsmodels'

For the change detector on the underlying trend we use another Kalman filter, this time a single period autoregression. We then consider the Malhalanobis distance between that filter output, using a Gumbel distribution to decide where the increase in distance likely indicates a change in trend. This implements the approach described by Lee & Roberts at https://www.robots.ox.ac.uk/~sjrob/Pubs/LeeRoberts_EVT.pdf

