JXAutoMLCompatibleModels 0.0.1

Models refactored to work with JXAutoML

JXAutoMLCompatibleModels

Author GitHub: https://github.com/TGChenZP

Please cite when using this package for research and other machine learning purposes

Introduction

This package wraps various model classes in an Sklearn style API, enabling its compatability with JXAutoML tuners.

Installation

pip install JXAutoMLCompatibleModels

Models

JXAutoMLCompatibleModels.Stacking.StackingClassifier_Pro [source]

class JXAutoMLCompatibleModels.Stacking.StackingClassifier_Pro(estimators_list, num_estimators, cv, n_jobs, passthrough, stack_method, verbose, final_estimator, **kwargs)

Parameters

Parameter & Type	Description
estimators (list of (str, estimator))	Base estimators to be stacked together. Each element is a tuple consisting of a string (name) and an estimator instance. An estimator can be set to ‘drop’ using set_params. Typically, these are classifiers, though regressors can be used for specific cases (e.g., ordinal regression).
final_estimator (estimator, default=None)	Classifier used to combine the base estimators. Defaults to LogisticRegression.
cv (int, cross-validation generator, iterable, or “prefit”, default=None)	Determines the cross-validation splitting strategy used in cross_val_predict to train final_estimator. Possible values include: - None: Default 5-fold cross-validation. - int: Specifies the number of folds in a (Stratified) KFold. - cross-validation generator: Object used as a cross-validation generator. - iterable: Yields train, test splits. - prefit: Assumes the estimators are prefit, and they will not be refitted. For integer/None inputs, StratifiedKFold is used for binary or multiclass classifiers; otherwise, KFold is used. Note that these splitters are instantiated with shuffle=False to ensure the same splits across calls. The ‘prefit’ option was added in version 1.1, assuming that all estimators are pre-fitted. The final_estimator_ is trained on the full training set predictions rather than cross-validated predictions, which risks overfitting if models are trained on the same data.
stack_method ({‘auto’, ‘predict_proba’, ‘decision_function’, ‘predict’}, default=auto)	Method called for each base estimator. Options include: - auto: Invokes, in order, predict_proba, decision_function, or predict. - predict_proba, decision_function, or predict: Specific method to be called. An error is raised if the method is not implemented by the estimator.
n_jobs (int, default=None)	The number of jobs to run in parallel during the fit of all estimators. None means 1, unless in a joblib.parallel_backend context. -1 means using all processors.
passthrough (bool, default=False)	When False, only the predictions of estimators will be used as training data for final_estimator. When True, the final_estimator is trained on the predictions as well as the original training data.
verbose (int, default=0)	Verbosity level.
**kwargs	Parameters of final_estimator can be defined through **kwargs and will be picked up during the call, allowing further customization of the final_estimator.

## JXAutoMLCompatibleModels.Stacking.StackingRegressor_Pro [source]

class JXAutoMLCompatibleModels.Stacking.StackingRegressor_Pro(estimators_list, num_estimators, cv, n_jobs, passthrough, verbose, final_estimator, **kwargs)

Parameters

Parameter & Type	Description
estimators (list of (str, estimator))	Base estimators to be stacked together. Each element is a tuple consisting of a string (name) and an estimator instance. An estimator can be set to ‘drop’ using set_params.
final_estimator (estimator, default=None)	Regressor used to combine the base estimators. Defaults to RidgeCV.
cv (int, cross-validation generator, iterable, or “prefit”, default=None)	Determines the cross-validation splitting strategy used in cross_val_predict to train final_estimator. Possible values include: - None: Default 5-fold cross-validation. - int: Specifies the number of folds in a (Stratified) KFold. - cross-validation generator: Object used as a cross-validation generator. - iterable: Yields train, test splits. - prefit: Assumes the estimators are prefit, and they will not be refitted. For integer/None inputs, StratifiedKFold is used for binary or multiclass classifiers; otherwise, KFold is used. Note that these splitters are instantiated with shuffle=False to ensure the same splits across calls. The ‘prefit’ option was added in version 1.1, assuming that all estimators are pre-fitted. The final_estimator_ is trained on the full training set predictions rather than cross-validated predictions, which risks overfitting if models are trained on the same data.
n_jobs (int, default=None)	The number of jobs to run in parallel during the fit of all estimators. None means 1, unless in a joblib.parallel_backend context. -1 means using all processors.
passthrough (bool, default=False)	When False, only the predictions of estimators will be used as training data for final_estimator. When True, the final_estimator is trained on the predictions as well as the original training data.
verbose (int, default=0)	Verbosity level.
**kwargs	Parameters of final_estimator can be defined through **kwargs and will be picked up during the call, allowing further customization of the final_estimator.

Usage Examples

StackingRegressor_Pro

from JXAutoMLCompatibleModels.Stacking import StackingRegressor_Pro
from sklearn.ensemble import RandomForestRegressor, GradientBoostingRegressor
from sklearn.svm import SVR
from sklearn.metrics import r2_score


X_reg, y_reg = make_regression(
        n_samples=100, n_features=5, noise=0.1, random_state=42)
X = pd.DataFrame(
    X_reg, columns=[f'feature_{i+1}' for i in range(X_reg.shape[1])])
y = pd.Series(y_reg, name='target')

model = StackingRegressor_Pro(
        estimators_list=[('gbr', GradientBoostingRegressor()), ('svr', SVR())],
        num_estimators=2,
        final_estimator_class=RandomForestRegressor,
        verbose=False,
        passthrough=True,
        n_jobs=-1,
        cv=3,
        n_estimators=50,
    )

model.fit(X, y)

print(r2_score(y, model.predict(X)))

StackingClassifier_Pro

from JXAutoMLCompatibleModels.Stacking import StackingClassifier_Pro
from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score

X_reg, y_reg = make_regression(
        n_samples=100, n_features=5, noise=0.1, random_state=42)
X = pd.DataFrame(
    X_reg, columns=[f'feature_{i+1}' for i in range(X_reg.shape[1])])
y = pd.Series(y_reg, name='target')

model = StackingClassifier_Pro(
        estimators_list=[
            ('gbr', GradientBoostingClassifier()), ('svc', SVC())],
        num_estimators=2,
        final_estimator_class=RandomForestClassifier,
        verbose=False,
        passthrough=True,
        stack_method='auto',
        n_jobs=-1,
        cv=3,
        n_estimators=50,
    )
    print("===Initialise Classifier complete===")

model.fit(X, y)

print(r2_score(y, model.predict(X)))