aeon: a Python toolkit for learning from time series

aeon: a Python toolkit for learning from time series

20 Jun 2024 | Matthew Middlehurst, Ali Ismail-Fawaz, Antoine Guillaume, Christopher Holder, David Guijo-Rubio, Guzal Bulatova, Leonidas Tsaprounis, Lukasz Mentel, Martin Walter, Patrick Schäfer, Anthony Bagnall
**aeon** is a unified Python 3 library designed for all machine learning tasks involving time series data. It includes modules for forecasting, classification, extrinsic regression, and clustering, along with various utilities, transformations, and distance measures tailored for time series data. aeon follows the scikit-learn API to facilitate easy integration with model selection and pipelines. The library supports a broad range of algorithms, including recent research advances, and integrates multiple packages through optional dependencies to maintain a minimal core framework. aeon is available under the 3-Clause BSD license and is distributed on GitHub. The authors of aeon are from various institutions, including the University of Southampton, Université de Haute-Alsace, ENSTA Paris, University of East Anglia, University of Córdoba, and Humboldt-Universität zu Berlin. The package is designed to be user-friendly, especially for those familiar with scikit-learn, and aims to unify diverse research fields and communities in time series machine learning (TSML). aeon's design is modular, with algorithms grouped by learning tasks. It uses object-oriented design and adheres to the scikit-learn estimator interface. The package supports Python 3.8 and later versions and includes core dependencies such as scikit-learn, numpy, and scipy. Optional dependencies include statsmodels, tensorflow, and tsfresh. Key features of aeon include: - **Forecasting**: Unified interface for popular forecasting tools and models. - **Classification, Clustering, and Regression**: Consistent *fit* and *predict* interfaces, compatible with scikit-learn. - **Transformations**: Objects that transform data from one representation to another, useful in pipelines. - **Experimental Modules**: Currently includes segmentation, anomaly detection, similarity search, and benchmarking. aeon is supported by the EPSRC and aims to be the most comprehensive toolkit for TSML, contributing to the development of the Python time series ecosystem.**aeon** is a unified Python 3 library designed for all machine learning tasks involving time series data. It includes modules for forecasting, classification, extrinsic regression, and clustering, along with various utilities, transformations, and distance measures tailored for time series data. aeon follows the scikit-learn API to facilitate easy integration with model selection and pipelines. The library supports a broad range of algorithms, including recent research advances, and integrates multiple packages through optional dependencies to maintain a minimal core framework. aeon is available under the 3-Clause BSD license and is distributed on GitHub. The authors of aeon are from various institutions, including the University of Southampton, Université de Haute-Alsace, ENSTA Paris, University of East Anglia, University of Córdoba, and Humboldt-Universität zu Berlin. The package is designed to be user-friendly, especially for those familiar with scikit-learn, and aims to unify diverse research fields and communities in time series machine learning (TSML). aeon's design is modular, with algorithms grouped by learning tasks. It uses object-oriented design and adheres to the scikit-learn estimator interface. The package supports Python 3.8 and later versions and includes core dependencies such as scikit-learn, numpy, and scipy. Optional dependencies include statsmodels, tensorflow, and tsfresh. Key features of aeon include: - **Forecasting**: Unified interface for popular forecasting tools and models. - **Classification, Clustering, and Regression**: Consistent *fit* and *predict* interfaces, compatible with scikit-learn. - **Transformations**: Objects that transform data from one representation to another, useful in pipelines. - **Experimental Modules**: Currently includes segmentation, anomaly detection, similarity search, and benchmarking. aeon is supported by the EPSRC and aims to be the most comprehensive toolkit for TSML, contributing to the development of the Python time series ecosystem.
Reach us at info@study.space