WeightedEnsembleClassifier
Weighted ensemble of classifiers with fittable ensemble weight.
Produces a probabilistic prediction which is the weighted average of predictions of individual classifiers. Classifier with name name has ensemble weight in weights_[name]. weights_ is fitted in fit, if weights is a scalar, otherwise fixed.
If weights is a scalar, empirical training loss is computed for each classifier. In this case, ensemble weights of classifier is empirical loss, to the power of weights (a scalar).
The evaluation for the empirical training loss can be selected through the metric and metric_type parameters.
The in-sample empirical training loss is computed in-sample or out-of-sample, depending on the cv parameter. None = in-sample; other = cross-validated oos.
Quickstart
from sktime.classification.ensemble import WeightedEnsembleClassifier
estimator = WeightedEnsembleClassifier(classifiers, weights=None, cv=None, metric=None, metric_type='point', random_state=None)Parameters(6)
- classifiersdict or None, default=None
- Parameters for the ShapeletTransformClassifier module. If None, uses the default parameters with a 2 hour transform contract.
- weightsfloat, or iterable of float, optional, default=None
if float, ensemble weight for classifier i will be train score to this power if iterable of float, must be equal length as classifiers
ensemble weight for classifier i will be weights[i]
if None, ensemble weights are equal (uniform average)
- cvNone, int, or sklearn cross-validation object, optional, default=None
determines whether in-sample or which cross-validated predictions used in fit None: predictions are in-sample, equivalent to fit(X, y).predict(X) cv: predictions are equivalent to fit(X_train, y_train).predict(X_test)
where multiple X_train, y_train, X_test are obtained from cv folds returned y is union over all test fold predictions cv test folds must be non-intersecting
- intequivalent to cv=KFold(cv, shuffle=True, random_state=x),
i.e., k-fold cross-validation predictions out-of-sample random_state x is taken from self if exists, otherwise x=None
- metricsklearn metric for computing training score, default=accuracy_score
- only used if weights is a float
- metric_typestr, one of “point” or “proba”, default=”point”
- type of sklearn metric, point prediction (“point”) or probabilistic (“proba”) if “point”, most probable class is passed as y_pred if “proba”, probability of most probable class is passed as y_pred
- random_stateint or None, default=None
- Seed for random number generation.
Examples
>>> from sktime.classification.dummy import DummyClassifier
>>> from sktime.classification.kernel_based import RocketClassifier
>>> from sktime.datasets import load_unit_test
>>> X_train, y_train = load_unit_test (split = "train")
>>> X_test, y_test = load_unit_test (split = "test")
>>> clf = WeightedEnsembleClassifier (
... [DummyClassifier (), RocketClassifier (num_kernels = 100)],
... weights = 2,
... )
>>> clf. fit (X_train, y_train) WeightedEnsembleClassifier(
... )
>>> y_pred = clf. predict (X_test)