carbatpy.models.components.surrogates

Created on Tue Jul 30 16:15:25 2024 structure to create Surrogate models 01.08.2024: for multilayer perceptrons (have proven themselves) needs a labeled DataFrame with training data

@author: welp

Attributes

mode

Classes

Surrogate

Module Contents

class carbatpy.models.components.surrogates.Surrogate(title)

title

train_surrogate(DF, features_list, targets_list, split=0.2, random_state=42, hypo='def_hyperparameter.yaml', verbose=False)

train MLP surrogate from dataframe with chosen features and targets, uses minmaxscaler and ‘r2’, ‘neg_root_mean_squared_error’ for scoring, the latter for refitting

Parameters:

• DF (TYPE) – DESCRIPTION.

• features_list (TYPE) – DESCRIPTION.

• targets_list (TYPE) – DESCRIPTION.

• split (TYPE, optional) – DESCRIPTION. The default is 0.2.

• random_state (TYPE, optional) – DESCRIPTION. The default is 42.

• hypo (TYPE, optional) – DESCRIPTION. The default is “def_hyperparameter.yaml”.

Returns:

• Y_test (TYPE DataFrame) – DESCRIPTION. test targets

• Y_pred (TYPE) – DESCRIPTION. predicted targets

save(path='default')

save model

Parameters:

path (TYPE, optional) – DESCRIPTION. The default is CARBATPY_RES_DIR

Return type:

None.

load(path)

predict(DF_new_x)

use existing surrogate to predict new data

Parameters:

DF_new_x (TYPE) – DESCRIPTION.

Raises:

ValueError –

DESCRIPTION. needs to contain features of model

DESCRIPTION. does not extrapolate

Returns:

• y (TYPE) – DESCRIPTION. target results

• DF_y (TYPE) – DESCRIPTION. target results

carbatpy.models.components.surrogates.mode = 1