tspy.forecasters module

main entry-point for creation of ForecastingModel

tspy.forecasters.anomaly_detector(confidence)

Build a AnomalyDetector model

Parameters

confidence: float

Returns

AnomalyDetector

tspy.forecasters.arima(error_horizon_length=1, use_full_error_history=True, force_model=False, min_training_data=- 1, p_min=0, p_max=- 1, d=- 1, q_min=0, q_max=- 1)

Build a ARIMA model

Parameters

error_horizon_length: int, optional (1)

use_full_error_history: bool, optional (True)

force_model: bool, optional (False)

min_training_data: int, optional (1)

p_min: int, optional (0)

p_max: int, optional (-1)

d: int, optional (-1)

q_min: int, optional (0)

q_max: int, optional (-1)

Returns

ForecastingModel

tspy.forecasters.arma(min_training_data=- 1, p_min=0, p_max=5, q_min=0, q_max=5)

Build a ARMA model

Parameters

min_training_data: int, optional (1)

p_min: int, optional (0)

p_max: int, optional (5)

q_min: int, optional (0)

q_max: int, optional (5)

Returns

ForecastingModel

tspy.forecasters.auto(min_training_data, error_history_length=1)

Build a AUTO model

Parameters

min_training_data: int

error_history_length: int, optional (1)

Returns

ForecastingModel

tspy.forecasters.bats(training_sample_size, box_cox_transform=False)

Build a BATS model

Parameters

training_sample_size: int

box_cox_transform: bool, optional

Returns

ForecastingModel

tspy.forecasters.hws(**kwargs)

Build a Holt-Winters model

Parameters

sample_per_season: int

initial_training_season:

Returns

ForecastingModel

tspy.forecasters.load(path)

tspy.forecasters.season_selector(sub_season_percent_delta=0.0, max_season_length=None)

Build a SeasonSelector model

Parameters

sub_season_percent_delta: float, optional (0.0)

max_season_length: int, optional (None)

Returns

SeasonSelector