Genetic Algorithm Python API Reference

This document provides comprehensive API reference for the genetic algorithm Python interfaces in the Ensemble Genetic Algorithm project.

---

Table of Contents

• GA Core Functions

• Fitness Evaluation

• Mutation Methods

• Weighting Methods

---

GA Core Functions

ml_grid.pipeline.ensemble_generator_ga.ensembleGenerator

Generates an initial ensemble for the genetic algorithm.

Function: ensembleGenerator

Creates a new ensemble individual by selecting base learners from the model list.

Module: ml_grid.pipeline.ensemble_generator_ga

See also: Configuration Guide for hyperparameter configuration options with local_param_dict.

Signature:

individual = ensembleGenerator(
    nb_val,
    ml_grid_object
)

See also: Pipeline API for data pipeline setup with ml_grid_object.

Parameters:

Parameter	Type	Description
nb_val	int	Number of base learners in the ensemble
ml_grid_object	Any	Experiment object containing model list and configuration

Returns: list

• Ensembles are represented as lists where individual[0] contains the base learners

---

ml_grid.pipeline.get_feature_selection_class_ga

Feature selection utilities for GA pipeline.

Function: feature_selection_methods_class.get_featured_selected_training_data

Selects features based on various importance methods (see Feature Selection for details).

Module: ml_grid.pipeline.get_feature_selection_class_ga

See also: Configuration Guide for n_features and feature_selection_method parameters, and Data_Workflow.md for feature selection details.

See also: Pipeline API for comprehensive feature selection utilities.

Signature:

X_train_selected, X_test_selected = feature_selection_methods_class(
    ml_grid_object
).get_featured_selected_training_data(method="anova")

Parameters:

Parameter	Type	Default	Description
method	str	"anova"	Feature selection method: “anova”, “randomforest”, etc.

Returns: Tuple[pd.DataFrame, pd.DataFrame]

• Selected features for training set

• Selected features for test set

---

Fitness Evaluation

ml_grid.pipeline.evaluate_methods_ga.get_y_pred_resolver

Resolves and generates predictions for ensemble evaluation (see Pipeline_API for GA pipeline overview).

Function: get_y_pred_resolver

Dispatches to the appropriate prediction method based on weighting configuration.

Module: ml_grid.pipeline.evaluate_methods_ga

Signature:

y_pred = get_y_pred_resolver(
    individual,
    ml_grid_object,
    valid=False
)

Parameters:

Parameter	Type	Default	Description
individual	List	Required	Ensemble configuration (DEAP individual format)
ml_grid_object	Any	Required	Experiment object with data splits
valid	bool	False	If True, predict on validation set

Returns: Union[List, np.ndarray]

• Final ensemble predictions

Prediction Methods:

The function dispatches based on weighted parameter in local_param_dict:

Weighted Type	Method Called
"unweighted"	get_unweighted_ensemble_predictions()
"de"	get_weighted_ensemble_prediction_de_y_pred_valid()
"ann"	get_y_pred_ann_torch_weighting()

---

ml_grid.pipeline.evaluate_methods_ga.evaluate_weighted_ensemble_auc

Main fitness evaluation function for genetic algorithm (see Pipeline_API for GA pipeline context).

Function: evaluate_weighted_ensemble_auc

Evaluates an individual (ensemble) and returns fitness score.

Module: ml_grid.pipeline.evaluate_methods_ga

Signature:

fitness = evaluate_weighted_ensemble_auc(
    individual,
    ml_grid_object
)

Parameters:

Parameter	Type	Description
individual	List	Ensemble to evaluate
ml_grid_object	Any	Experiment object with data and configuration

Returns: Tuple[float]

• Single-element tuple containing fitness score (AUC or diversity-penalized AUC)

Evaluation Process:

1. Generate predictions using get_y_pred_resolver()

2. Calculate performance metrics:

   • AUC: Receiver Operating Characteristic Area Under Curve

   • MCC: Matthews Correlation Coefficient

   • F1, Precision, Recall, Accuracy

3. Measure ensemble diversity using measure_binary_vector_diversity()

4. Apply diversity penalty if configured (div_p > 0 - see Configuration Guide for GA parameters)

5. Log results to CSV file

Diversity Penalty:

When div_p > 0, the fitness score is adjusted:

auc_div, mcc_div = apply_diversity_penalty(
    auc, mcc, diversity_metric, diversity_params
)

Diversity Parameters (from config - see Configuration Guide):

• penalty_method: "linear", "quadratic", "exponential", or "threshold"

• penalty_strength: User-specified magnitude (default: 0.3)

• min_score_factor: Minimum score multiplier (default: 0.1)

---

ml_grid.pipeline.evaluate_methods_ga.normalize

Normalizes weight vectors.

Function: normalize

L1-normalizes a vector of weights.

Module: ml_grid.pipeline.evaluate_methods_ga

Signature:

normalized = normalize(weights)

Parameters:

Parameter	Type	Description
weights	np.ndarray	Array of weights to normalize

Returns: np.ndarray

• L1-normalized weight vector (sum of absolute values = 1)

---

ml_grid.pipeline.evaluate_methods_ga.measure_binary_vector_diversity

Measures diversity between prediction vectors.

Function: measure_binary_vector_diversity

Calculates how diverse the predictions are across ensemble members.

Module: ml_grid.pipeline.evaluate_methods_ga

Signature:

diversity = measure_binary_vector_diversity(
    ensemble,
    metric="jaccard"
)

Parameters:

Parameter	Type	Default	Description
ensemble	List	Required	Ensemble with base learners
metric	str	"jaccard"	Distance metric for diversity

Returns: float

• Mean pairwise distance between prediction vectors

Supported Metrics:

• "jaccard": Jaccard similarity distance

• "hamming": Hamming distance

• Other scipy.spatial.distance metrics

---

Mutation Methods

ml_grid.pipeline.mutate_methods.baseLearnerGenerator

Generates a random base learner (see Base Learner Interface for generator interface details, and Pipeline_API.md for GA pipeline context).

Function: baseLearnerGenerator

Creates a new base learner by randomly selecting from the model list.

Module: ml_grid.pipeline.mutate_methods

Signature:

new_learner = baseLearnerGenerator(ml_grid_object)

Parameters:

Parameter	Type	Description
ml_grid_object	Any	Experiment object with modelFuncList configuration

Returns: Any

• Instance of a randomly selected model generator

---

ml_grid.pipeline.mutate_methods.mutateEnsemble

Mutates an ensemble by replacing one base learner (see Pipeline_API for GA pipeline context).

Function: mutateEnsemble

Performs genetic mutation on an ensemble individual.

Module: ml_grid.pipeline.mutate_methods

Signature:

mutated_individual = mutateEnsemble(
    individual,
    ml_grid_object
)

Parameters:

Parameter	Type	Description
individual	List	Ensemble to mutate (individual[0] contains base learners)
ml_grid_object	Any	Experiment object with model configuration

Returns: List

• Mutated individual with one base learner replaced

Mutation Process:

1. Randomly select a position in the ensemble

2. Remove the base learner at that position

3. Generate a new random baselearner using baseLearnerGenerator()

4. Append the new learner to complete mutation

---

Weighting Methods

Unweighted Ensemble Prediction

See also: Pipeline_API.md for GA pipeline workflow.

Function: ml_grid.ga_functions.ga_unweighted.get_unweighted_ensemble_predictions

Generates predictions by majority voting (mode).

Module: ml_grid.ga_functions.ga_unweighted

Signature:

predictions = get_unweighted_ensemble_predictions(
    best,
    ml_grid_object,
    valid=False
)

Parameters:

Parameter	Type	Default	Description
best	List	Required	Best ensemble configuration
ml_grid_object	Any	Required	Experiment object with data splits
valid	bool	False	Use validation set if True

Returns: List

• Final predictions via mode of individual model predictions

---

DE Weighted Ensemble Prediction

See also: Pipeline_API.md for pipeline workflow and Configuration_Guide.md for weighted prediction options ("de" method).

Function: ml_grid.ga_functions.ga_de_weight_method.get_weighted_ensemble_prediction_de_y_pred_valid

Generates predictions using Differential Evolution weighted ensemble.

Module: ml_grid.ga_functions.ga_de_weight_method

Signature:

predictions = get_weighted_ensemble_prediction_de_y_pred_valid(
    ensemble,
    weights,
    ml_grid_object,
    valid=False
)

Parameters:

Parameter	Type	Default	Description
ensemble	List	Required	Ensemble configuration
weights	np.ndarray	Required	Learner weights from optimization
ml_grid_object	Any	Required	Experiment object
valid	bool	False	Use validation set if True

Returns: Union[List, np.ndarray]

• Weighted ensemble predictions

---

DE Weight Ensemble Finder

See also: Pipeline_API.md for pipeline workflow and Configuration_Guide.md for weight optimization parameters.

Function: ml_grid.ga_functions.ga_ensemble_weight_finder_de.find_ensemble_weights_de

Finds optimal weights for ensemble using Differential Evolution.

Module: ml_grid.ga_functions.ga_ensemble_weight_finder_de

Signature:

weights = find_ensemble_weights_de(
    ensemble,
    ml_grid_object,
    valid=False
)

Parameters:

Parameter	Type	Default	Description
ensemble	List	Required	Ensemble to optimize weights for
ml_grid_object	Any	Required	Experiment object
valid	bool	False	Optimize on validation set if True

Returns: np.ndarray -Optimized weight vector

---

ANN Weighted Ensemble Prediction

See also: Pipeline_API.md for pipeline workflow and Configuration_Guide.md for weighted prediction options ("ann" method).

Function: ml_grid.ga_functions.ga_ann_weight_methods.get_y_pred_ann_torch_weighting

Generates predictions using neural network-based ensemble weighting.

Module: ml_grid.ga_functions.ga_ann_weight_methods

Signature:

predictions = get_y_pred_ann_torch_weighting(
    ensemble,
    ml_grid_object,
    valid=False
)

Parameters:

Parameter	Type	Default	Description
ensemble	List	Required	Ensemble configuration
ml_grid_object	Any	Required	Experiment object
valid	bool	False	Use validation set if True

Returns: Union[List, np.ndarray]

• Neural network-weighted predictions

---

Utility Functions

ml_grid.util.ensemble_diversity_methods

Diversity measurement and penalty application.

Function: measure_diversity_wrapper

Wrapper for ensemble diversity measurement.

Module: ml_grid.util.ensemble_diversity_methods

Signature:

diversity = measure_diversity_wrapper(
    individual,
    method="comprehensive"
)

Function: apply_diversity_penalty

Applies diversity-based penalty to fitness scores.

Module: ml_grid.util.ensemble_diversity_methods

Signature:

auc_penalized, mcc_penalized = apply_diversity_penalty(
    auc_score,
    mcc_score,
    diversity_metric,
    params=diversity_params
)

---

Version Information

This API documentation corresponds to version v1.0+ of the ensemble_genetic_algorithm package.

Python Requirement: Python >=3.12

To check your installed version:

pip show ensemble_genetic_algorithm