Skip to content

Welcome to Yohou-Nixtla's documentation

Yohou-Nixtla brings the power of Nixtla's forecasting backends (StatsForecast and NeuralForecast) into the Yohou ecosystem. Each backend is wrapped as a scikit-learn-compatible Yohou forecaster with full support for fit, predict, observe, and rewind, so you can use classical statistical models and deep learning architectures through a single unified API.

Powered by Nixtla

Under the hood, Yohou-Nixtla delegates to Nixtla libraries. Data is automatically converted between Yohou's polars wide-format and Nixtla's pandas long-format, so you never need to wrangle DataFrames yourself.

  • Get Started in 5 Minutes

    Install Yohou-Nixtla and produce your first forecast

    Getting Started

  • Need Help?

    Find answers to common questions and troubleshooting tips.

    Troubleshooting

  • Learn the Concepts

    Understand the forecasting backends, data conversion, and panel data

    Concepts

  • See It In Action

    Compare statistical and neural forecasters on real data

    Examples

Key Features

  • 15 forecasters, one API: 10 statistical models (AutoARIMA, AutoETS, Holt-Winters, ...) and 5 neural architectures (NBEATS, NHITS, PatchTST, ...) all sharing fit / predict / observe / rewind.

  • Yohou and Scikit-Learn-compatible: Every forecaster supports clone, get_params, and set_params. Use Yohou's GridSearchCV for time series hyperparameter search.

  • Panel data out of the box: Name columns with the __ separator (sales__store_1) and Yohou-Nixtla fits each group independently in a single call.

  • Automatic data conversion: Polars wide-format DataFrames are converted to Nixtla's pandas long-format transparently on every fit and predict call.

  • Exogenous features: Pass external regressors through X with optional feature_transformer for automatic scaling and preprocessing.

  • Minimal boilerplate: Adding a new Nixtla model is a three-line class. No glue code, no manual DataFrame wrangling.

License

This project is licensed under the terms of the Apache-2.0 License.

Acknowledgements

This project is maintained by stateful-y, an ML consultancy specializing in time series data science & engineering. If you're interested in collaborating or learning more about our services, please visit our website.

Made by stateful-y