skchange

skchange provies sktime-compatible change detection and changepoint-based anomaly detection algorithms. Methods implement the annotator interface of sktime.

A playground for now.

Quickstart

Changepoint detection / time series segmentation

from skchange.change_detectors.moscore import Moscore
from skchange.datasets.generate import generate_teeth_data

df = generate_teeth_data(n_segments=10, segment_length=50, mean=5, random_state=1)
detector = Moscore(bandwidth=10, fmt="sparse")
detector.fit_predict(df)
>>>
0     49
1     99
2    149
3    199
4    249
5    299
6    349
7    399
8    449
Name: changepoints, dtype: int32

Multivariate anomaly detection

from skchange.anomaly_detectors.mvcapa import Mvcapa
from skchange.datasets.generate import generate_teeth_data

df = generate_teeth_data(
    n_segments=5,
    segment_length=50,
    p=10,
    mean=10,
    affected_proportion=0.2,
    random_state=2,
)
detector = Mvcapa(collective_penalty="sparse", fmt="sparse")
detector.fit_predict(df)
>>>
   start  end components
0     50   99     [0, 1]
1    150  199     [0, 1]

Installation

pip install git+https://github.com/NorskRegnesentral/skchange

Dependencies

• pandas >= 1.3
• numpy >= 1.19
• numba >= 0.56
• sktime >= 0.24

You also need Python >= 3.8.

License

skchange is a free and open-source software licensed under the BSD 3-clause license.