brew package

Subpackages

• brew.combination package
  • Submodules
  • brew.combination.combiner module
  • brew.combination.rules module
  • Module contents
• brew.generation package
  • Submodules
  • brew.generation.bagging module
  • brew.generation.base module
  • brew.generation.random_subspace module
  • brew.generation.smote_bagging module
  • Module contents
• brew.metrics package
  • Subpackages
    • brew.metrics.diversity package
      • Submodules
      • brew.metrics.diversity.base module
      • brew.metrics.diversity.non_paired module
      • brew.metrics.diversity.paired module
      • Module contents
  • Submodules
  • brew.metrics.evaluation module
  • Module contents
• brew.preprocessing package
  • Submodules
  • brew.preprocessing.smote module
  • Module contents
• brew.selection package
  • Subpackages
    • brew.selection.dynamic package
      • Submodules
      • brew.selection.dynamic.base module
      • brew.selection.dynamic.knora module
      • brew.selection.dynamic.lca module
      • brew.selection.dynamic.ola module
      • Module contents
    • brew.selection.pruning package
      • Submodules
      • brew.selection.pruning.epic module
      • Module contents
  • Module contents
• brew.utils package
  • Submodules
  • brew.utils.data module
  • Module contents

Submodules

brew.base module

class brew.base.BrewClassifier(classifier=None, transformer=None)

Bases: object

fit(X, y)

predict(X)

predict_proba(X)

class brew.base.Ensemble(classifiers=None)

Bases: object

Class that represents a collection of classifiers.

The Ensemble class serves as a wrapper for a list of classifiers, besides providing a simple way to calculate the output of all the classifiers in the ensemble.

`classifiers`

list – Stores all classifiers in the ensemble.

`yval`

array-like, shape = [indeterminated] – Labels of the validation set.

`knn`

sklearn KNeighborsClassifier, – Classifier used to find neighborhood.

Examples

>>> import numpy as np
>>> from sklearn.tree import DecisionTreeClassifier
>>>
>>> from brew.base import Ensemble
>>>
>>> X = np.array([[-1, 0], [-0.8, 1], [-0.8, -1], [-0.5, 0],
                  [0.5, 0], [1, 0], [0.8, 1], [0.8, -1]])
>>> y = np.array([1, 1, 1, 2, 1, 2, 2, 2])
>>>
>>> dt1 = DecisionTreeClassifier()
>>> dt2 = DecisionTreeClassifier()
>>>
>>> dt1.fit(X, y)
>>> dt2.fit(X, y)
>>>
>>> ens = Ensemble(classifiers=[dt1, dt2])

add(classifier)

add_classifiers(classifiers)

add_ensemble(ensemble)

fit(X, y)

warning: this fit overrides previous generated base classifiers!

get_classes()

in_agreement(x)

output(X, mode='votes')

Returns the output of all classifiers packed in a numpy array.

This method calculates the output of each classifier, and stores them in a array-like shape. The specific shape and the meaning of each element is defined by argument mode.

(1) ‘labels’: each classifier will return a single label prediction for each sample in X, therefore the ensemble output will be a 2d-array of shape (n_samples, n_classifiers), with elements being the class labels.

(2) ‘probs’: each classifier will return the posterior probabilities of each class (i.e. instead of returning a single choice it will return the probabilities of each class label being the right one). The ensemble output will be a 3d-array with shape (n_samples, n_classes, n_classifiers), with each element being the probability of a specific class label being right on a given sample according to one the classifiers. This mode can be used with any combination rule.

(3) ‘votes’: each classifier will return votes for each class label i.e. a binary representation, where the chosen class label will have one vote and the other labels will have zero votes. The ensemble output will be a binary 3d-array with shape (n_samples, n_classes, n_classifiers), with the elements being the votes. This mode is mainly used in combining the classifiers output by using majority vote rule.

Parameters:

• X (array-like, shape = [n_samples, n_features]) – The test input samples.
• mode (string, optional(default='labels')) – The type of output given by each classifier. ‘labels’ | ‘probs’ | ‘votes’

output_simple(X)

class brew.base.EnsembleClassifier(ensemble=None, selector=None, combiner=None)

Bases: object

fit(X, y)

predict(X)

predict_proba(X)

score(X, y, sample_weight=None)

class brew.base.FeatureSubsamplingTransformer(features=None)

Bases: brew.base.Transformer

apply(X)

class brew.base.Transformer

Bases: object

apply(X)

brew.base.oracle(ensemble, X, y_true, metric=<function auc_score>)

brew.base.single_best(ensemble, X, y_true, metric=<function auc_score>)

brew.base.transform2votes(output, n_classes)

Module contents

class brew.Ensemble(classifiers=None)

Bases: object

Class that represents a collection of classifiers.

The Ensemble class serves as a wrapper for a list of classifiers, besides providing a simple way to calculate the output of all the classifiers in the ensemble.

`classifiers`

list – Stores all classifiers in the ensemble.

`yval`

array-like, shape = [indeterminated] – Labels of the validation set.

`knn`

sklearn KNeighborsClassifier, – Classifier used to find neighborhood.

Examples

>>> import numpy as np
>>> from sklearn.tree import DecisionTreeClassifier
>>>
>>> from brew.base import Ensemble
>>>
>>> X = np.array([[-1, 0], [-0.8, 1], [-0.8, -1], [-0.5, 0],
                  [0.5, 0], [1, 0], [0.8, 1], [0.8, -1]])
>>> y = np.array([1, 1, 1, 2, 1, 2, 2, 2])
>>>
>>> dt1 = DecisionTreeClassifier()
>>> dt2 = DecisionTreeClassifier()
>>>
>>> dt1.fit(X, y)
>>> dt2.fit(X, y)
>>>
>>> ens = Ensemble(classifiers=[dt1, dt2])

add(classifier)

add_classifiers(classifiers)

add_ensemble(ensemble)

fit(X, y)

warning: this fit overrides previous generated base classifiers!

get_classes()

in_agreement(x)

output(X, mode='votes')

Returns the output of all classifiers packed in a numpy array.

This method calculates the output of each classifier, and stores them in a array-like shape. The specific shape and the meaning of each element is defined by argument mode.

(1) ‘labels’: each classifier will return a single label prediction for each sample in X, therefore the ensemble output will be a 2d-array of shape (n_samples, n_classifiers), with elements being the class labels.

(2) ‘probs’: each classifier will return the posterior probabilities of each class (i.e. instead of returning a single choice it will return the probabilities of each class label being the right one). The ensemble output will be a 3d-array with shape (n_samples, n_classes, n_classifiers), with each element being the probability of a specific class label being right on a given sample according to one the classifiers. This mode can be used with any combination rule.

(3) ‘votes’: each classifier will return votes for each class label i.e. a binary representation, where the chosen class label will have one vote and the other labels will have zero votes. The ensemble output will be a binary 3d-array with shape (n_samples, n_classes, n_classifiers), with the elements being the votes. This mode is mainly used in combining the classifiers output by using majority vote rule.

Parameters:

• X (array-like, shape = [n_samples, n_features]) – The test input samples.
• mode (string, optional(default='labels')) – The type of output given by each classifier. ‘labels’ | ‘probs’ | ‘votes’

output_simple(X)

class brew.EnsembleClassifier(ensemble=None, selector=None, combiner=None)

Bases: object

fit(X, y)

predict(X)

predict_proba(X)

score(X, y, sample_weight=None)