------------------------------------------------------------------------ source ### ets ``` python def ets( target_col:str, # Name of the target variable column. trend:Optional[str]=None, # Trend component type. damped_trend:bool=False, # Whether to damp the trend. Only meaningful when ``trend`` is not `None`. seasonal:Optional[str]=None, # Seasonal component type. seasonal_periods:Optional[int]=None, # Number of periods in a complete seasonal cycle — e.g. 12 for monthly data with an annual cycle. Required when ``seasonal`` is not ``None``. initialization_method:Optional[str]='estimated', # How to initialise the recursions. When `"known"` is chosen, `initial_level` (and `initial_trend` / `initial_seasonal` where applicable) must also be provided. initial_level:Optional[float]=None, # Initial level value. Required when `initialization_method=`"known"`. initial_trend:Optional[float]=None, # Initial trend value. Required when `initialization_method=`"known"` and the model has a trend component. initial_seasonal:Optional[list]=None, # Initial seasonal factors (length `seasonal_periods` or ``seasonal_periods - 1``). Required when ``initialization_method="known"`` and the model is seasonal. bounds:Optional[dict]=None, # Parameter bounds passed to ``ExponentialSmoothing``, e.g. `{"smoothing_level": (0, 1)}`. dates:NoneType=None, # Datetime index for the series. Inferred automatically when `endog` is a Pandas object with a `DatetimeIndex`. freq:Optional[str]=None, # Frequency of the time series (e.g. ``"M"``, ``"D"``). Optional when `dates` is provided. missing:str='none', # How to handle ``NaN`` values in the input series. optimized:bool=True, # Estimate smoothing parameters by maximising the log-likelihood. smoothing_level:Optional[float]=None, # Fixed alpha value. When set, this value is used directly and not optimised. smoothing_trend:Optional[float]=None, # Fixed beta value. Only used when the model has a trend component. smoothing_seasonal:Optional[float]=None, # Fixed gamma value. Only used when the model is seasonal. damping_trend:Optional[float]=None, # Fixed phi (damping) value. Only used when ``damped_trend=True``. remove_bias:bool=False, # Remove bias from forecast values by enforcing that the mean residual is zero. start_params:NoneType=None, # Starting parameter values for the optimiser. method:Optional[str]=None, # Optimisation method — one of `"L-BFGS-B"` (default), `"TNC"`, `"SLSQP"`, `"Powell"`, `"trust-constr"`, `"basinhopping"` (alias `"bh"`), or `"least_squares"` (alias `"ls"`). minimize_kwargs:Optional[dict]=None, # Extra keyword arguments forwarded to the chosen SciPy minimiser. use_brute:bool=True, # Search for good starting values with a brute-force grid search before running the main optimiser. box_cox:Union[bool, float]=False, # Whether to apply Box-Cox transformation to the target variable. If a float value is provided, it will be used as the lambda parameter for the Box-Cox transformation. If True, the lambda parameter will be estimated from the data. box_cox_biasadj:bool=False, # Bias adjustment when inverting the manual Box-Cox on forecasts. fit_kwargs:Optional[dict]=None, # Any additional keyword arguments forwarded verbatim to `ExponentialSmoothing.fit`. )->None: # Fitted model object with a ``forecast`` method for making predictions and properties for information criteria scores (AIC, BIC, etc.) ``` *Holt-Winters Exponential Smoothing forecaster.* A thin wrapper around `statsmodels.tsa.holtwinters.ExponentialSmoothing`.

	Type	Default	Details
target_col	str		Name of the target variable column.
trend	Optional[str]	None	Trend component type.
damped_trend	bool	False	Whether to damp the trend. Only meaningful when trend is not None.
seasonal	Optional[str]	None	Seasonal component type.
seasonal_periods	Optional[int]	None	Number of periods in a complete seasonal cycle — e.g. 12 for monthly data with an annual cycle. Required when seasonal is not None.
initialization_method	Optional[str]	estimated	How to initialise the recursions. When "known" is chosen, initial_level (and initial_trend / initial_seasonal where applicable) must also be provided.
initial_level	Optional[float]	None	Initial level value. Required when initialization_method=“known”. | | initial_trend | Optional[float] | None | Initial trend value. Required wheninitialization_method="known" and the model has a trend component.
initial_seasonal	Optional[list]	None	Initial seasonal factors (length seasonal_periods or seasonal_periods - 1). Required when initialization_method="known" and the model is seasonal.
bounds	Optional[dict]	None	Parameter bounds passed to ExponentialSmoothing, e.g. {"smoothing_level": (0, 1)}.
dates	NoneType	None	Datetime index for the series. Inferred automatically when endog is a Pandas object with a DatetimeIndex.
freq	Optional[str]	None	Frequency of the time series (e.g. "M", "D"). Optional when dates is provided.
missing	str	none	How to handle NaN values in the input series.
optimized	bool	True	Estimate smoothing parameters by maximising the log-likelihood.
smoothing_level	Optional[float]	None	Fixed alpha value. When set, this value is used directly and not optimised.
smoothing_trend	Optional[float]	None	Fixed beta value. Only used when the model has a trend component.
smoothing_seasonal	Optional[float]	None	Fixed gamma value. Only used when the model is seasonal.
damping_trend	Optional[float]	None	Fixed phi (damping) value. Only used when damped_trend=True.
remove_bias	bool	False	Remove bias from forecast values by enforcing that the mean residual is zero.
start_params	NoneType	None	Starting parameter values for the optimiser.
method	Optional[str]	None	Optimisation method — one of "L-BFGS-B" (default), "TNC", "SLSQP", "Powell", "trust-constr", "basinhopping" (alias "bh"), or "least_squares" (alias "ls").
minimize_kwargs	Optional[dict]	None	Extra keyword arguments forwarded to the chosen SciPy minimiser.
use_brute	bool	True	Search for good starting values with a brute-force grid search before running the main optimiser.
box_cox	Union[bool, float]	False	Whether to apply Box-Cox transformation to the target variable. If a float value is provided, it will be used as the lambda parameter for the Box-Cox transformation. If True, the lambda parameter will be estimated from the data.
box_cox_biasadj	bool	False	Bias adjustment when inverting the manual Box-Cox on forecasts.
fit_kwargs	Optional[dict]	None	Any additional keyword arguments forwarded verbatim to ExponentialSmoothing.fit.
Returns	None		Fitted model object with a forecast method for making predictions and properties for information criteria scores (AIC, BIC, etc.)

------------------------------------------------------------------------ source ### ets.fit ``` python def fit( df:pd.DataFrame, # Training DataFrame containing the target column )->None: ``` *Fit `ExponentialSmoothing` to the training data.*

	Type	Details
df	pd.DataFrame	Training DataFrame containing the target column
Returns	None

------------------------------------------------------------------------ source ### ets.forecast ``` python def forecast( H:int, # Forecast horizon. exog:Optional[pd.DataFrame]=None, # Accepted for API consistency with other models but silently ignored — ETS forecasts do not use exogenous variables. )->np.ndarray: # Forecast values of length `H`. ``` *Multi-step forecast.*

	Type	Default	Details
H	int		Forecast horizon.
exog	Optional[pd.DataFrame]	None	Accepted for API consistency with other models but silently ignored — ETS forecasts do not use exogenous variables.
Returns	np.ndarray		Forecast values of length H.

------------------------------------------------------------------------ source ### ets.cross_validate ``` python def cross_validate( df:pd.DataFrame, # Full dataset cv_split:int, # Number of CV folds. test_size:int, # Test window size per fold. metrics:List[Callable], # Metric functions (e.g. ``[MAE, RMSE]``) used to evaluate forecast accuracy across folds. Call ``.cv_summary()`` after cross-validation to retrieve the aggregated scores. step_size:int=1, # Step size to advance the test window each fold. h_split_point:Optional[int]=None, # Split the test window into two sub-horizons for separate short- and long-term evaluation. )->Tuple[pd.DataFrame, pd.DataFrame]: # Summary DataFrame with mean metric scores across folds, and (optionally) a fold-level DataFrame with true vs. predicted values for each fold. ``` *Run time-series cross-validation.*

	Type	Default	Details
df	pd.DataFrame		Full dataset
cv_split	int		Number of CV folds.
test_size	int		Test window size per fold.
metrics	List[Callable]		Metric functions (e.g. [MAE, RMSE]) used to evaluate forecast accuracy across folds. Call .cv_summary() after cross-validation to retrieve the aggregated scores.
step_size	int	1	Step size to advance the test window each fold.
h_split_point	Optional[int]	None	Split the test window into two sub-horizons for separate short- and long-term evaluation.
Returns	Tuple[pd.DataFrame, pd.DataFrame]		Summary DataFrame with mean metric scores across folds, and (optionally) a fold-level DataFrame with true vs. predicted values for each fold.