Classifier

SklearnClassifierPipeline

API-Referenz Quellcode

Multivariate Predict proba Unequal length Missing values

Pipeline of transformers and a classifier.

The SklearnClassifierPipeline chains transformers and an single classifier.: Similar to ClassifierPipeline, but uses a tabular sklearn classifier.
The pipeline is constructed with a list of sktime transformers, plus a classifier,: i.e., transformers following the BaseTransformer interface, classifier follows the scikit-learn classifier interface.
The transformer list can be unnamed - a simple list of transformers -: or string named - a list of pairs of string, estimator.

For a list of transformers trafo1, trafo2, …, trafoN and a classifier clf,

the pipeline behaves as follows:

fit(X, y) - changes styte by running trafo1.fit_transform on X,: them trafo2.fit_transform on the output of trafo1.fit_transform, etc sequentially, with trafo[i] receiving the output of trafo[i-1], and then running clf.fit with X the output of trafo[N] converted to numpy, and y identical with the input to self.fit. X is converted to numpyflat mtype if X is of Panel scitype; X is converted to numpy2D mtype if X is of Table scitype.

predict(X) - result is of executing trafo1.transform, trafo2.transform, etc

with trafo[i].transform input = output of trafo[i-1].transform, then running clf.predict on the numpy converted output of trafoN.transform, and returning the output of clf.predict. Output of trasfoN.transform is converted to numpy, as in fit.

predict_proba(X) - result is of executing trafo1.transform, trafo2.transform,

etc, with trafo[i].transform input = output of trafo[i-1].transform, then running clf.predict_proba on the output of trafoN.transform, and returning the output of clf.predict_proba. Output of trasfoN.transform is converted to numpy, as in fit.

get_params, set_params uses sklearn compatible nesting interface

if list is unnamed, names are generated as names of classes if names are non-unique, f"_{str(i)}" is appended to each name string

where i is the total count of occurrence of a non-unique string inside the list of names leading up to it (inclusive)

SklearnClassifierPipeline can also be created by using the magic multiplication

between sktime transformers and sklearn classifiers,: and my_trafo1, my_trafo2 inherit from BaseTransformer, then, for instance, my_trafo1 * my_trafo2 * my_clf will result in the same object as obtained from the constructor SklearnClassifierPipeline(classifier=my_clf, transformers=[t1, t2])
magic multiplication can also be used with (str, transformer) pairs,: as long as one element in the chain is a transformer

Schnellstart

python

from sktime.classification.compose import SklearnClassifierPipeline

estimator = SklearnClassifierPipeline(classifier, transformers)

Parameter(2)

classifiersklearn classifier, i.e., inheriting from sklearn ClassifierMixin: this is a “blueprint” classifier, state does not change when fit is called
transformerslist of sktime transformers, or: list of tuples (str, transformer) of sktime transformers these are “blueprint” transformers, states do not change when fit is called

Beispiele

>>> from sklearn.neighbors import KNeighborsClassifier
>>> from sktime.transformations.exponent import ExponentTransformer
>>> from sktime.transformations.summarize import SummaryTransformer
>>> from sktime.datasets import load_unit_test
>>> from sktime.classification.compose import SklearnClassifierPipeline
>>> X_train, y_train = load_unit_test (split = "train")
>>> X_test, y_test = load_unit_test (split = "test")
>>> t1 = ExponentTransformer ()
>>> t2 = SummaryTransformer ()
>>> pipeline = SklearnClassifierPipeline (KNeighborsClassifier (), [t1, t2 ])
>>> pipeline = pipeline. fit (X_train, y_train)
>>> y_pred = pipeline. predict (X_test) Alternative construction via dunder method:
>>> pipeline = t1 * t2 * KNeighborsClassifier ()