autots.evaluator.feature_detector.components package

Submodules

autots.evaluator.feature_detector.components.anomalies module

AnomalyMixin for TimeSeriesFeatureDetector.

class autots.evaluator.feature_detector.components.anomalies.AnomalyMixin

Bases: object

Mixin providing anomaly detection, noise analysis, and regime changepoint detection.

autots.evaluator.feature_detector.components.decomposition module

DecompositionMixin for TimeSeriesFeatureDetector.

class autots.evaluator.feature_detector.components.decomposition.DecompositionMixin

Bases: object

Mixin providing initial decomposition (rough seasonality removal).

autots.evaluator.feature_detector.components.holidays module

HolidayMixin for TimeSeriesFeatureDetector.

class autots.evaluator.feature_detector.components.holidays.HolidayMixin

Bases: object

Mixin providing holiday detection, coefficient solving, and parameter sanitization.

autots.evaluator.feature_detector.components.joint_trend module

JointTrendMixin: opt-in joint hinge/step trend estimation for TimeSeriesFeatureDetector.

Enabled with changepoint_params={'method': 'joint_hinge_step', ...}. PELT-l2 segments a ramp into a level staircase and the legacy level-shift check accepts most random candidates; here slope changes (hinges) and level shifts (steps) are selected and sized in one least-squares model with a BIC stop and AR(1)-adjusted step t-tests. Also hosts the opt-in t-test level-shift validator (level_shift_validation={'method': 't_test'}) usable with any changepoint method.

class autots.evaluator.feature_detector.components.joint_trend.JointTrendMixin

Bases: object

Joint hinge/step trend selection and t-test level-shift validation.

autots.evaluator.feature_detector.components.joint_trend.mutate_joint_trend_changepoint_params(params, rng=None, n_keys=2)

Resample a few method_params within the declared space.

Generic numeric jitter would push booleans/ints (seasonal_prior, backfit_passes) and the BIC multiplier outside meaningful ranges.

autots.evaluator.feature_detector.components.joint_trend.sample_joint_trend_changepoint_params(rng=None)

Random changepoint_params for the joint method (for get_new_params / optimizers).

autots.evaluator.feature_detector.components.seasonality module

SeasonalityMixin for TimeSeriesFeatureDetector.

class autots.evaluator.feature_detector.components.seasonality.SeasonalityMixin

Bases: object

Mixin providing seasonality fitting, strength estimation, and changepoint detection.

autots.evaluator.feature_detector.components.trend module

TrendMixin for TimeSeriesFeatureDetector.

class autots.evaluator.feature_detector.components.trend.TrendMixin

Bases: object

Mixin providing trend changepoint detection, level shift detection, and slope computation.

Module contents