Scientific type of the object.

Typing tag for all objects in sktime.

  • String name: "object_type"

  • Public metadata tag

  • Values: string or list of strings

  • Example: "forecaster"

  • Example 2: ["transformer", "clusterer"] (polymorphic object)

  • Default: "object"

In sktime, every object has a scientific type (scitype), determining the type of object and unified interface, e.g., forecaster, time series classifier, time series regressor.

The object_type tag of an object is a string, or list of strings, specifying the scitpye of the object. For instance, a forecaster has scitype “forecaster”.

In case of a list, the object is polymorphic, and can assume (class), or simultaneously satisfy different interfaces (object).

Valid scitypes are defined in sktime.registry.BASE_CLASS_SCITYPE_LIST, or sktime.registry.BASE_CLASS_REGISTER.

The full list of scitypes in the current version is:

  • "object": object

  • "estimator": estimator = object with fit

  • "aligner": time series aligner or sequence aligner

  • "classifier": time series classifier

  • "clusterer": time series clusterer

  • "early_classifier": early time series classifier

  • "forecaster": forecaster

  • "metric": performance metric

  • "network": deep learning network

  • "param_est": parameter fitting estimator

  • "regressor": time series regressor

  • "series-annotator": time series annotator

  • "splitter": time series splitter

  • "transformer": time series transformer

  • "transformer-pairwise": pairwise transformer for tabular data, distance or kernel

  • "transformer-pairwise-panel": pairwise transformer for panel data, distance or kernel

  • "distribution": pandas-like probability distribution