kats.models.holtwinters module

The Holt Winters model is a time series forecast model that applies exponential smoothing three times, it serves as the extension of the simple exponential smoothing forecast model.

More details about the different exponential smoothing model can be found here: https://en.wikipedia.org/wiki/Exponential_smoothing In this module we adopt the Holt Winters model implementation from the statsmodels package, full details can be found as follows: https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing.html We rewrite the corresponding API to accommodate the Kats development style

class kats.models.holtwinters.HoltWintersModel(data: kats.consts.TimeSeriesData, params: kats.models.holtwinters.HoltWintersParams)

Bases: Generic[kats.models.model.ParamsType]

Model class for the HoltWinters model

data

kats.consts.TimeSeriesData, the input historical time series data from TimeSeriesData

params

The HoltWinters model parameters from HoltWintersParams

fit(**kwargs) → None

Fit the model with the specified input parameters

static get_parameter_search_space() → List[Dict[str, Any]]

Get default HoltWinters parameter search space.

Parameters

None –

Returns

A dictionary with the default HoltWinters parameter search space

plot()

Plot forecast results from the HoltWinters model

predict(steps: int, include_history: bool = False, **kwargs) → pandas.core.frame.DataFrame

Predict with fitted HoltWinters model

If the alpha keyword argument is specified, an empirical confidence interval is computed through a K-fold cross validation and a linear regression model, the forecast outcome will include a confidence interval there; otherwise no confidence interval is included in the final forecast. Please refer to the ‘emp_confidence_int’ module for full detailed implementation of the empirical confidence interval computation

Parameters

• steps –

• include_history –

Returns

A pd.DataFrame with the forecast and confidence interval (if empirical confidence interval calculation is triggered)

class kats.models.holtwinters.HoltWintersParams(trend: str = 'add', damped: bool = False, seasonal: Optional[str] = None, seasonal_periods: Optional[int] = None)

Bases: kats.consts.Params

Parameter class for the HoltWinters model

Not all parameters from the statsmodels API have been implemented here, the full list of the parameter can be found: https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing.html

trend

Optional; A string that specifies the type of trend component. Can be ‘add’ and ‘mul’ or ‘additive’ and ‘multiplicative’. Default is ‘add’

damped

Optional; A boolean indicates whether the trend should be damped or not. Default is False

seasonal

Optional; A string that specifies the type of seasonal component Can be ‘add’ and ‘mul’ or ‘additive’ and ‘multiplicative’. Default is None

seasonal_periods

Optional; An integer that specifies the period for the seasonal component, e.g. 4 for quarterly data and 7 for weekly seasonality of daily data. Default is None

validate_params()

Validate the types and values of the input parameters :param None:

Returns

None