Supervised learning

In supervised learning, we are given a collection of labeled time series and the goal is to produce an estimator that as accurately as possible can map previously unseen time series to the correct label.

Shapelet trees

For classification, wildboar includes two types of shapelet trees. Both trees are based on randomly sampling shapelets. ShapeletTreeClassifier samples shapelets randomly and ExtraShapeletTreeClassifier also sample the distance threshold randomly. Both ShapeletTreeRegressor and ExtraShapeletTreeRegressor are available.

Shapelet forest

Shapelet forests, implemented in ShapeletForestClassifier and ShapeletForestRegressor, construct ensembles of shapelet tree classifiers or regressors respectively. For a large variety of tasks, these estimators are excellent baseline methods.

Proximity forest

Proximity Forest is an ensemble of highly randomized Proximity Trees. Whereas conventional decision trees branch on attribute values, and shapelet trees on distance thresholds, Proximity Trees is k-branching tree that branches on proximity of time series to one of k pivot time series.

Embedding methods

RocketClassifier and RocketRegressor uses a random convolutional embdding to represent time series and fit a ridge regression model to the representation. For benchmark tasks, this transformation and estimator configuration often give state-of-the-art predictive performance.