SplineTrendForecaster

Forecast time series data with a spline trend.

Quickstart

from sktime.forecasting.trend import SplineTrendForecaster

estimator = SplineTrendForecaster(regressor=None, n_knots=5, degree=1, knots='uniform', extrapolation='constant', with_intercept=True)

Parameters(7)

regressorsklearn regressor estimator object, default=None

Define the regression model type. If not set, defaults to sklearn.linear_model.LinearRegression.

n_knotsint, default=5

Number of knots of the splines if knots is one of {‘uniform’, ‘quantile’}. Must be at least 2. Ignored if knots is array-like.

degreeint, default=1

Degree of the splines (1 for linear, 2 for quadratic, etc.).

n_knotsint, default=4

Number of knots for the spline transformation.

knots{‘uniform’, ‘quantile’}or array-like of shape (n_knots, n_features),

default=’uniform’ Determines knot positions such that first knot <= features <= last knot.

• ‘uniform’: n_knots are distributed uniformly between the

min and max values of the features. - ‘quantile’: n_knots are distributed uniformly along the quantiles of the features. - array-like: Specifies sorted knot positions, including the boundary knots. Internally, additional knots are added before the first knot and after the last knot based on the spline degree.

extrapolation{‘error’, ‘constant’, ‘linear’, ‘continue’, ‘periodic’},

default=’constant’ Determines how to handle values outside the min and max values of the training features:

• ‘error’: Raises a ValueError.

• ‘constant’: Uses the spline value at the minimum or maximum feature as

constant extrapolation. - ‘linear’: Applies linear extrapolation. - ‘continue’: Extrapolates as is (equivalent to extrapolate=True in scipy.interpolate.BSpline). - ‘periodic’: Uses periodic splines with a periodicity equal to the distance between the first and last knot, enforcing equal function values and derivatives at these knots.

with_interceptbool, default=True

If True, includes a feature in which all polynomial powers are zero (i.e., a column of ones, acting as an intercept term in a linear model).

Examples

>>> from sktime.datasets import load_airline
>>> from sktime.forecasting.trend import SplineTrendForecaster
>>> y = load_airline ()
>>> forecaster = SplineTrendForecaster (
... n_knots = 5,
... degree = 2,
... knots = "uniform",
... extrapolation = "constant"
... )
>>> forecaster. fit (y) SplineTrendForecaster(
... )
>>> y_pred = forecaster. predict (fh = [1, 2, 3 ])