environmentaltools.temporal.regression

environmentaltools.temporal.regression(base_train, target_train, base_pred, method='rbf-multiquadric', num=100, smooth=1, optimize=True, eps=1, optimizer='local')

Performs regression using various interpolation and machine learning methods.

This function supports multiple regression approaches including interpolation methods (linear, cubic, nearest), radial basis functions (RBF), and Gaussian processes (GP).

Parameters:

• base_train (pd.DataFrame or np.ndarray) – Training input features (predictors)

• target_train (pd.Series or np.ndarray) – Training target values (response)

• base_pred (pd.DataFrame or np.ndarray) – Test input features for prediction

• method (str, optional) –

  Regression method. Defaults to ‘rbf-multiquadric’. Available methods:

  • Interpolation: ‘linear’, ‘nearest’, ‘cubic’

  • RBF: ‘rbf-multiquadric’, ‘rbf-inverse’, ‘rbf-gaussian’, ‘rbf-linear’, ‘rbf-cubic’, ‘rbf-quintic’, ‘rbf-thin_plate’

  • Gaussian Process: ‘gp-rbf’, ‘gp-exponential’, ‘gp-quadratic’, ‘gp-white’

• num (int, optional) – Number of points for RBF optimization. Defaults to 100.

• smooth (float, optional) – Smoothing parameter for RBF methods. Defaults to 1.

• optimize (bool, optional) – Whether to optimize RBF epsilon parameter. Defaults to True.

• eps (float, optional) – Manual epsilon parameter for RBF (used if optimize=False). Defaults to 1.

• optimizer (str, optional) – Optimization method (‘local’ or other). Defaults to ‘local’.

Returns:

Predicted values for input base_pred

Return type:

np.ndarray

Raises:

ValueError – If the specified method is not implemented

Examples

>>> predictions = regression(
...     base_train=train_features,
...     target_train=train_targets,
...     base_pred=test_features,
...     method='rbf-multiquadric',
...     optimize=True
... )