Forecasting

The aeon.forecasting module contains algorithms for forecasting.

All forecasters in aeon can be listed using the aeon.utils.discovery .all_estimators function using type_filter="forecaster", optionally filtered by tags. Valid tags for forecasters can be found with aeon.utils.tags .all_tags_for_estimator function with the argument "forecaster".

Forecasting Models

BaseForecaster(horizon, axis)	Abstract base class for time series forecasters.
NaiveForecaster([strategy, seasonal_period, ...])	Naive forecaster with multiple strategies and flexible horizon.
RegressionForecaster(window[, horizon, ...])	Regression based forecasting.

Statistical Models

ARIMA([p, d, q, use_constant, iterations])	AutoRegressive Integrated Moving Average (ARIMA) forecaster.
AutoARIMA([max_p, max_d, max_q])	AutoRegressive Integrated Moving Average (ARIMA) forecaster.
ETS([error_type, trend_type, ...])	Exponential Smoothing (ETS) forecaster.
AutoETS()	Automatic Exponential Smoothing forecaster.
TAR([threshold, delay, ar_order])	Threshold Autoregressive (TAR) [1] forecaster with fixed parameters.
AutoTAR([threshold, delay, ar_order, ...])	Threshold Autoregressive (AutoTAR) forecaster with fast threshold search.
Theta([theta, weight])	Classical Theta forecaster.
TVP(window[, horizon, var, beta_var])	Time-Varying Parameter (TVP) Forecaster using Kalman filter as described in [1].

Deep Learning Models

BaseDeepForecaster(window[, horizon, ...])	Base class for deep learning forecasters in aeon.
DeepARForecaster(window[, horizon, ...])	A deep learning forecaster using DeepAR architecture.
TCNForecaster(window[, horizon, batch_size, ...])	A deep learning forecaster using Temporal Convolutional Network (TCN).

Machine Learning Models

SETAR([lag])	Self-Exciting Threshold AutoRegressive (SETAR) forecaster with 2 regimes.
SETARForest([lag, n_estimators, ...])	SETAR-Forest: Bagging + random subspace ensemble of SETAR-Tree base learners.
SETARTree([lag, max_depth, ...])	SETAR-Tree: A tree algorithm for global time series forecasting.