aeon

As time series analysis has grown critical across manufacturing IoT, healthcare wearables, and financial forecasting, aeon consolidates the scattered landscape of specialized algorithmsROCKET, HIVE-COTE, MiniRocket, and others from the UCR/UEA benchmark literaturebehind a consistent API that lets practitioners swap methods without restructuring code.

The library covers the full time series ML workflow: transformation (feature extraction, dimensionality reduction, segmentation), supervised learning (classification and regression with both classical and deep learning approaches), unsupervised analysis (clustering, anomaly detection), and similarity search.

Aeon's compatibility with scikit-learn pipelines enables these time series-specific estimators to integrate naturally with the broader Python ML ecosystem, including cross-validation, grid search, and pipeline composition tools that practitioners already use daily.

Data scientists working on predictive maintenance from vibration sensor data, healthcare signal classification from ECG and EEG recordings, and financial time series forecasting use aeon to access state-of-the-art time series methods without implementing algorithms from academic papers.

The package is the successor to sktime's time series classification component, inheriting its extensive algorithm library while focusing development on this problem domain.

Research groups contribute new algorithm implementations directly, making aeon a living reference for the current state of the art in time series machine learning.