wildboar

wildboar is a Python module for temporal machine learning and fast distance computations built on top of scikit-learn and numpy distributed under the BSD 3-Clause license.

It is currently maintained by Isak Samsten

Features

Data	Classification	Regression	Explainability	Metric	Unsupervised	Dimension selection	Outlier
Repositories	ShapeletForestClassifier	ShapeletForestRegressor	ShapeletForestCounterfactual	UCR-suite	ShapeletForestTransform	DimensionVarianceThreshold	IsolationShapeletForest
Classification (wildboar/ucr)	ExtraShapeletTreesClassifier	ExtraShapeletTreesRegressor	KNearestCounterfactual	MASS	RandomShapeletEmbedding	SelectDimensionPercentile
Regression (wildboar/tsereg)	RocketTreeClassifier	RocketRegressor	PrototypeCounterfactual	DTW	RocketTransform	SelectDimensionTopK
Outlier detection (wildboar/outlier:easy)	RocketClassifier	RandomShapeletRegressor	IntervalImportance	DDTW	IntervalTransform	SelectDimensionSignificance
	RandomShapeletClassifier	RocketTreeRegressor	ShapeletImportance	WDTW	FeatureTransform
	RocketForestClassifier	RocketForestRegressor		MSM	MatrixProfileTransform
	IntervalTreeClassifier	IntervalTreeRegressor		TWE	FlossSegmenter
	IntervalForestClassifier	IntervalForestRegressor		LCSS	Motif discovery
	ProximityTreeClassifier	CastorRegressor		ERP	SAX
	ProximityForestClassifier			EDR	PAA
	HydraClassifier			ADTW	MatrixProfileTransform
	KNeighborsClassifier				HydraTransform
	ElasticEnsembleClassifier				KMeans with (W)DTW support
	DilatedShapeletClassifier				KMedoids
	CastorClassifier				DilatedShapeletTransform
					CastorTransform
					QuantTransform

See the documentation for examples.

Installation

Binaries

wildboar is available through pip and can be installed with:

pip install wildboar

Universal binaries are compiled for Python 3.9, 3.10, 3.11 and 3.12 running on GNU/Linux, Windows and macOS.

Compilation

If you already have a working installation of numpy, scikit-learn, scipy and cython, compiling and installing wildboar is as simple as:

pip install .

To install the requirements, use:

pip install -r requirements.txt

For complete instructions see the documentation

Usage

from wildboar.ensemble import ShapeletForestClassifier
from wildboar.datasets import load_dataset
x_train, x_test, y_train, y_test = load_dataset("GunPoint", merge_train_test=False)
c = ShapeletForestClassifier()
c.fit(x_train, y_train)
c.score(x_test, y_test)

The User guide includes more detailed usage instructions.

Changelog

The changelog records a history of notable changes to wildboar.

Development

Contributions are welcome! The developer's guide has detailed information about contributing code and more!

In short, pull requests should:

• be well motivated
• be formatted using Black
• add relevant tests
• add relevant documentation

Source code

You can check the latest sources with the command:

git clone https://github.com/wildboar-foundation/wildboar

Documentation

• HTML documentation: https://wildboar.dev

Citation

If you use wildboar in a scientific publication, I would appreciate citations to the paper:

• Karlsson, I., Papapetrou, P. Boström, H., 2016. Generalized Random Shapelet Forests. In the Data Mining and Knowledge Discovery Journal

  • ShapeletForestClassifier

• Isak Samsten, 2020. isaksamsten/wildboar: wildboar. Zenodo. doi:10.5281/zenodo.4264063

• Karlsson, I., Rebane, J., Papapetrou, P. et al. Locally and globally explainable time series tweaking. Knowl Inf Syst 62, 1671–1700 (2020)

  • ShapeletForestCounterfactual
  • KNearestCounterfactual