Unified API
Provides a consistent interface for multiple time series learning tasks, allowing users to build, fit, apply, and validate models with the same conventions.
Dedicated Time Series Algorithms
Includes algorithms specifically designed for time series analysis rather than adapted from general-purpose methods.
Composite Model Building
Supports pipelines with transformations, ensembles, hyperparameter tuning, and task reduction to create complex models.
Scikit-learn-like Interface
Enables users familiar with scikit-learn to switch between models without changing code preparation or execution.
Hierarchical Forecasting Support
Allows application of different forecasting models at various levels of data aggregation.
Fair Model Assessment and Benchmarking
Provides tools to build, inspect, and validate models while avoiding common pitfalls in time series evaluation.