tsfuse

tsfuse.construct(X, y, task='auto', transformers='full', max_depth=2, return_graph=False, **kwargs)

Construct features for a labeled time series dataset X, y

Parameters

• X (dict(str, Collection)) – Time series data.

• y (array-like) – Labels. Since each window has a single label, the length should be equal to the number of windows in the given time series data.

• task ({'classification', 'regression', 'auto'}, default: 'auto') – Machine learning task: detected automatically by default.

• transformers ({'minimal', 'full'}, default: 'full') – Feature construction settings: ‘minimal’ uses a minimal set of simple statistical transformers and ‘full’ the complete set of transformers.

• max_depth (int, default: 2) – Maximum computation graph depth.

• return_graph (bool, default: False) – Return computation graph.

Returns

• features (pandas.DataFrame) – Constructed features.

• graph (Graph) – Computation graph that computes the constructed features. Only returned if return_graph == True