Drift detection

../../../../../../_images/MLPro-OA-Drift-Detectors_class_diagram.drawio.png

Ver. 0.2.0 (2025-03-03)

This module provides templates for drift detection to be used in the context of online adaptivity.

class mlpro.oa.streams.tasks.driftdetectors.basics.DriftDetector(p_name: str = None, p_range_max=1, p_ada: bool = True, p_duplicate_data: bool = False, p_visualize: bool = False, p_logging=True, p_drift_buffer_size: int = 100, **p_kwargs)

Bases: OAStreamTask

Base class for online anomaly detectors. It raises an event whenever the beginning or the end of a drift is detected. Please describe in child classes which event classes are used. Always use the _raise_drift_event() method when raising an event.

Parameters:

  • p_name (str) – Optional name of the task. Default is None.

  • p_range_max (int) – Maximum range of asynchonicity. See class Range. Default is Range.C_RANGE_PROCESS.

  • p_ada (bool) – Boolean switch for adaptivitiy. Default = True.

  • p_duplicate_data (bool) – If True, instances will be duplicated before processing. Default = False.

  • p_visualize (bool) – Boolean switch for visualisation. Default = False.

  • p_logging – Log level (see constants of class Log). Default: Log.C_LOG_ALL

  • p_drift_buffer_size (int = 100) – Size of the internal drift buffer self.drifts. Default = 100.

  • p_kwargs (dict) – Further optional named parameters.

C_TYPE = 'Drift Detector'
C_PLOT_ACTIVE: bool = True
C_PLOT_STANDALONE: bool = False
_get_next_drift_id()

Methd that returns the id of the next drift.

Returns:

drift_id

Return type:

int

_buffer_drift(p_drift: Drift)

Method to be used internally to add a new drift object. Please use as part of your algorithm.

Parameters:

p_drift (Drift) – Drift object to be added.

_remove_drift(p_drift: Drift)

Method to remove an existing drift object. Please use as part of your algorithm.

Parameters:

p_drift (Drift) – Drift object to be removed.

_raise_drift_event(p_drift: Drift, p_buffer: bool = True)

Specialized method to raise drift events.

Parameters:

p_drift (Drift) – Drift event object to be raised.

init_plot(p_figure: Figure = None, p_plot_settings: PlotSettings = None)

Initializes the plot functionalities of the class.

Parameters:

  • p_figure (Matplotlib.figure.Figure, optional) – Optional MatPlotLib host figure, where the plot shall be embedded. The default is None.

  • p_plot_settings (PlotSettings) – Optional plot settings. If None, the default view is plotted (see attribute C_PLOT_DEFAULT_VIEW).

update_plot(p_inst: Dict[int, Tuple[str, Instance]] = None, **p_kwargs)

Specialized definition of method update_plot() of class mlpro.bf.plot.Plottable.

Parameters:

  • p_inst (InstDict) – Instances to be plotted.

  • p_kwargs (dict) – Further optional plot parameters.

remove_plot(p_refresh: bool = True)

” Removes the plot and optionally refreshes the display.

Parameters:

p_refresh (bool = True) – On True the display is refreshed after removal

_renormalize(p_normalizer)

Internal renormalization of all buffered drifts. See method OATask.renormalize_on_event() for further information.

Parameters:

p_normalizer (Normalizer) – Normalizer object to be applied on task-specific

Ver. 0.3.0 (2025-02-19)

This module provides a template class for types of data drift to be used in drift detection algorithms.

class mlpro.oa.streams.tasks.driftdetectors.drifts.basics.Drift(p_drift_status: bool, p_id: int = 0, p_tstamp: datetime = None, p_visualize: bool = False, p_raising_object: object = None, **p_kwargs)

Bases: Id, Event, Plottable, Renormalizable

This is the base class for drift events which can be raised by drift detectors when the beginning or end of a drift is detected.

Parameters:

  • p_drift_status (bool) – Determines whether a new drift starts (True) or an existing drift ends (False).

  • p_id (int) – Drift ID. Default value = 0.

  • p_tstamp (datetime) – Time stamp of drift detection. Default = None.

  • p_visualize (bool) – Boolean switch for visualisation. Default = False.

  • p_raising_object (object) – Reference of the object raised. Default = None.

  • **p_kwargs – Further optional keyword arguments.

event_id

Event id to be used when raising a drift event object. It is a string consisting of the class name and one of the postfixes ‘(ON)’, ‘(OFF)’ depending on the drift status.

Type:

str

C_PLOT_STANDALONE: bool = False
_get_drift_status() → bool
_get_event_id() → str
property drift_status: bool
property event_id: str