capability__multivariate#

capability__multivariate()[source]#

Capability: the estimator can handle multivariate time series.

String name: "capability:multivariate"
Public capability tag
Values: boolean, True / False
Example: True
Default: False
Alias: univariate-only (transformations, note: boolean is inverted)
Alias: univariate-metric (performance metrics, note: boolean is inverted)

If the tag is True, the estimator can handle multivariate time series, for its main input data, i.e., the X parameter in fit of classifiers, regressors, clusterers, ordinary transformers, and pairwise transformers.

If the tag is False, the estimator can only handle univariate time series natively. Depending on the type of object, multivariate time series may be valid inputs, in this case the estimator will broadcast to variables (transformers, forecasters), or raise an error (others).

This condition is specific to the main input data representation, target data (e.g., classifier or transformation y) are not considered.

The condition is also specific to the data type used, in terms of how being “multivariate” is represented. For instance, a pandas based time series specification is considered multivariate if it has more than one column.