MODULE REFERENCE
timeseries
anomx.detection.detectors.timeseriesNative component-wise decisions for deterministic anomaly-score series.
IQRDetector
IQRDetector(scale: Bounds = 1.5)Bases: QuantileDetector
Learn Tukey fences Q1 - scale*IQR and Q3 + scale*IQR.
scale is nonnegative (default 1.5) and may be supplied per component. With constant training scores, any different value lies outside the fence.
| Parameter | Type | Default |
|---|---|---|
scale | Bounds | 1.5 |
Members
component_keyattribute
component_key = 'iqr'Inherited members
fitQuantileDetectordetectQuantileDetectorfit_detectQuantileDetectoris_component_abstractBaseComponentget_component_typeBaseComponentget_component_keyBaseComponentget_component_nameBaseComponentget_component_docsBaseComponentget_component_descriptionBaseComponentget_component_iconBaseComponentget_component_image_pathBaseComponentget_component_statusBaseComponentget_component_default_configBaseComponentget_component_config_schemaBaseComponentget_component_code_versionBaseComponentget_component_source_pathBaseComponentget_component_import_pathBaseComponentget_component_capabilitiesBaseComponentget_component_signatureBaseComponentget_component_parametersBaseComponentget_configBaseComponentQuantileDetector
QuantileDetector(low_quantile: Bounds = None, high_quantile: Bounds = 0.99)Bases: AdaptiveThresholdDetector
Learn lower/upper per-component quantiles from explicit training scores.
Unlike the legacy QuantileThresholdDetector, calibration is separate from detection, so evaluation data never changes its own threshold. Quantiles can be scalars or one probability per component; use None to disable either side. Fitting on a sequence pools observations while preserving component order. Probabilities must lie in [0, 1].
| Parameter | Type | Default |
|---|---|---|
low_quantile | Bounds | None |
high_quantile | Bounds | 0.99 |
Members
component_keyattribute
component_key = 'quantile'capabilitiesattribute
capabilities = ComponentCapabilities(supports_sequences=True, supports_unordered_samples=False, requires_training=True)fitmethod
fit(series: TimeSeries | Sequence[TimeSeries]) -> QuantileDetectorFit thresholds using training scores only.
| Parameter | Type | Default |
|---|---|---|
series | TimeSeries | Sequence[TimeSeries] | required |
Returns QuantileDetector
detectmethod
detect(series: TimeSeries | Sequence[TimeSeries]) -> TimeSeries | list[TimeSeries]Apply the fixed training thresholds to new observations.
| Parameter | Type | Default |
|---|---|---|
series | TimeSeries | Sequence[TimeSeries] | required |
Returns TimeSeries | list[TimeSeries]
fit_detectmethod
fit_detect(series: SeriesLike) -> TimeSeries | list[TimeSeries]Fit and label the same input; use separate data when evaluating performance.
| Parameter | Type | Default |
|---|---|---|
series | SeriesLike | required |
Returns TimeSeries | list[TimeSeries]
Inherited members
is_component_abstractBaseComponentget_component_typeBaseComponentget_component_keyBaseComponentget_component_nameBaseComponentget_component_docsBaseComponentget_component_descriptionBaseComponentget_component_iconBaseComponentget_component_image_pathBaseComponentget_component_statusBaseComponentget_component_default_configBaseComponentget_component_config_schemaBaseComponentget_component_code_versionBaseComponentget_component_source_pathBaseComponentget_component_import_pathBaseComponentget_component_capabilitiesBaseComponentget_component_signatureBaseComponentget_component_parametersBaseComponentget_configBaseComponentTimeSeriesThresholdDetector
TimeSeriesThresholdDetector(low_threshold: Bounds = None, high_threshold: Bounds = None)Bases: StaticThresholdDetector
Flag scores strictly below or above explicit component-wise bounds.
At least one of low_threshold and high_threshold is required. Each can be a scalar or one value per component. Equality is in bounds. Returns a binary TimeSeries (1 anomalous, 0 normal), retaining coordinates. This differs from the older table-oriented ThresholdDetector, which adds a column to a record table and uses an absolute-value threshold.
| Parameter | Type | Default |
|---|---|---|
low_threshold | Bounds | None |
high_threshold | Bounds | None |
Members
component_keyattribute
component_key = 'timeseries_threshold'capabilitiesattribute
capabilities = ComponentCapabilities(supports_sequences=True, supports_unordered_samples=False, requires_training=False)detectmethod
detect(series: TimeSeries | Sequence[TimeSeries]) -> TimeSeries | list[TimeSeries]Apply configured bounds to one or several independent score series.
| Parameter | Type | Default |
|---|---|---|
series | TimeSeries | Sequence[TimeSeries] | required |
Returns TimeSeries | list[TimeSeries]
Inherited members
is_component_abstractBaseComponentget_component_typeBaseComponentget_component_keyBaseComponentget_component_nameBaseComponentget_component_docsBaseComponentget_component_descriptionBaseComponentget_component_iconBaseComponentget_component_image_pathBaseComponentget_component_statusBaseComponentget_component_default_configBaseComponentget_component_config_schemaBaseComponentget_component_code_versionBaseComponentget_component_source_pathBaseComponentget_component_import_pathBaseComponentget_component_capabilitiesBaseComponentget_component_signatureBaseComponentget_component_parametersBaseComponentget_configBaseComponent