kats.detectors.detector module

Defines the base class for detectors.

class kats.detectors.detector.Detector(data: kats.consts.TimeSeriesData)

Bases: abc.ABC

Base detector class to be inherited by specific detectors

data

The input time series data from TimeSeriesData

remover(interpolate: bool = False) → kats.consts.TimeSeriesData

Remove the outlier in time series

Parameters

interpolate – Optional; bool; interpolate the outlier

Returns

A TimeSeriesData with outlier removed.

class kats.detectors.detector.DetectorModel(serialized_model: Optional[bytes])

Bases: abc.ABC

Base Detector model class to be inherited by specific detectors. A DetectorModel keeps the state of the Detector, and implements the incremental model training.

The usage of the DetectorModel is (replace DetectorModel with the proper child class)

model = DetectorModel(serialized_model) model.fit(new_data, …)

# the model may be saved through model.serialize() call # the model may be loaded again through model = DetectorModel(serialized_model)

result = model.predict(data, …)