swectral.modelcombiners.RegressorToClassifier#

class swectral.modelcombiners.RegressorToClassifier(regressor, proba_func='softmax')[source]#

Wrap a sklearn-style regressor into a sklearn-style classifier using one-hot encoding and probability regression.

This class converts a regressor that predicts continuous outputs into a classifier by one-hot encoding the targets and applying a probability conversion function to the regressor outputs.

Attributes:

regressorobject

A regressor implementing fit(X, y) and predict(X).

proba_funcstr or Callable, optional

Function to convert raw regressor outputs to probabilities. Choose between:

'softmax' : row-wise softmax for single-label, mutually exclusive classes

'sigmoid' : per-class sigmoid for multi-label problems

Callable : a custom probability function.

Default is 'softmax'.

encoderOneHotEncoder

Encoder for transforming class labels to one-hot vectors.

classes_numpy.ndarray of shape (n_classes,) or None

Array of class labels seen during fitting.

regressors_list of object or None

Fitted regressors. Contains a single regressor if multi-output regression is supported, otherwise one regressor per class.

Methods

fit(X, y)	Fit the regressor(s) on one-hot encoded class targets.
predict_proba(X)	Predict class probabilities for each sample.
predict(X)	Predict class labels for each sample.

Examples

>>> import numpy as np
>>> from sklearn.linear_model import LinearRegression
>>> X = np.array([[1], [2], [3], [4]])
>>> y = np.array([0, 1, 0, 1])

>>> reg = LinearRegression()

>>> clf = RegressorToClassifier(regressor=reg, proba_func='softmax')
>>> clf.fit(X, y)

>>> clf.predict(X)

>>> clf.predict_proba(X)

__init__(regressor, proba_func='softmax')[source]#

Methods

`__init__`(regressor[, proba_func])
`fit`(X, y)	Fit the regressor on one-hot encoded targets.
`predict`(X)	Predict class labels.
`predict_proba`(X)	Predict class probabilities.