pyccea.evaluation package

Submodules

pyccea.evaluation.wrapper module

class pyccea.evaluation.wrapper.WrapperEvaluation(task: str, model_type: str, eval_function: str, eval_mode: str, n_classes: int = None)

Bases: object

Evaluate selected features based on the predictive performance of a machine learning model.

Attributes:

model_evaluatorobject of one of the metrics classes

Responsible for computing performance metrics to evaluate models.

base_modelsklearn model object

Model that has not been fitted. Works as a template to avoid multiple model initializations. As each model evaluates a subset of features (individual), the base model is copied and fitted for each individual.

modelsklearn model object

Model that has been fitted to evaluate the current individual.

estimatorslist of sklearn model objects

Estimators used in the current evaluation. It is one when ‘eval_mode’ is set to “hold_out” and k when ‘eval_mode’ is set to “k_fold” or “leave_one_out”.

Methods

evaluate(solution, data)

Evaluate an individual represented by a complete solution through the predictive performance of a machine learning model.

eval_modes = ['hold_out', 'k_fold', 'leave_one_out']

evaluate(solution: ndarray, data: DataLoader) → dict

Evaluate an individual represented by a complete solution through the predictive performance of a machine learning model.

Parameters:

solutionnp.ndarray

Solution represented by a binary n-dimensional array, where n is the number of features.

dataDataLoader

Container with process data and training and test sets.

Returns:

: dict

Evaluation metrics.

metrics = {'classification': <class 'pyccea.utils.metrics.ClassificationMetrics'>, 'regression': <class 'pyccea.utils.metrics.RegressionMetrics'>}

models = {'classification': <class 'pyccea.utils.models.ClassificationModel'>, 'regression': <class 'pyccea.utils.models.RegressionModel'>}