Skip to content
stateful-yPublic

About

Time-series forecasting built on scikit-learn, with composable forecasters, transformers and cross-validation.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

7 stars

Watchers

1 watching

Forks

Latest commit

 

History

298 Commits

Folders and files

Yohou

Python Version License PyPI Version Conda Version codecov

What is Yohou?

Yohou is a time series forecasting framework built on top of Scikit-Learn's ecosystem. It provides a unified interface for building, extending, and comparing any forecasting model, from sklearn-native reductions to statistical models or deep learning integrations and hyperparameter optimization workflows. All models share a consistent API with native DataFrame support, Scikit-Learn-based compositions, and first-class cross-validation.

What are the features of Yohou?

  • Polars-native: All data flows use polars.DataFrame with a mandatory "time" column. No pandas required.
  • Sklearn-compatible: Standard fit/predict API with a consistent interface across all forecaster types.
  • Reduction forecasting: Wrap any Scikit-Learn regressor (Ridge, XGBRegressor, ...) and Yohou handles windowing, tabularization, and recursive prediction via PointReductionForecaster, IntervalReductionForecaster, and ClassProbaReductionForecaster.
  • Point, interval, and class-probability forecasting: From naive baselines to conformal prediction intervals (SplitConformalForecaster) and calibrated class-probability distributions.
  • Panel data: Prefix columns with group__ and all forecasters, transformers, and metrics operate across groups automatically. Use ColumnForecaster or LocalPanelForecaster for per-group models.
  • Incremental observation: Call observe() to feed new data, rewind() to roll back state, and observe_predict() to fast-forward and forecast in one step without retraining.
  • Stateful transformers: All transformers implement fit/observe/rewind and participate fully in incremental forecasting pipelines, enabling correct state management across training and deployment.
  • Composable pipelines: Chain trend, seasonality, and residual forecasters with DecompositionPipeline, or build feature pipelines with FeaturePipeline, FeatureUnion, and ColumnTransformer.
  • Cross-validation and model selection: Temporal splitters (ExpandingWindowSplitter, SlidingWindowSplitter) and GridSearchCV/RandomizedSearchCV designed for time series with no data leakage.
  • Metrics: Point, interval, and class-probability scorers with stepwise, vintagewise, componentwise, and groupwise aggregation.

How to install Yohou?

Install the Yohou package using pip:

pip install yohou

or using uv:

uv pip install yohou

or using conda:

conda install -c conda-forge yohou

or using mamba:

mamba install -c conda-forge yohou

or alternatively, add yohou to your requirements.txt or pyproject.toml file.

How to get started with Yohou?

1. Load data and split

Yohou datasets are fetched from Monash/Zenodo and return a Bunch with a .frame attribute (a Polars DataFrame with a "time" column).

from yohou.datasets import fetch_sunspot

bunch = fetch_sunspot()
y = bunch.frame
y_train, y_test = y[:-30], y[-30:]

2. Fit a forecaster

Wrap an sklearn regressor in a PointReductionForecaster with preprocessing pipelines.

from sklearn.linear_model import Ridge

from yohou.compose import FeaturePipeline
from yohou.point import PointReductionForecaster
from yohou.preprocessing import LagTransformer
from yohou.stationarity import SeasonalDifferencing

forecaster = PointReductionForecaster(
    estimator=Ridge(alpha=10),
    target_transformer=FeaturePipeline([
        ("diff", SeasonalDifferencing(seasonality=27)),
    ]),
    actual_transformer=FeaturePipeline([
        ("lag", LagTransformer(lag=[1, 2, 3, 27])),
    ]),
)
forecaster.fit(y_train, forecasting_horizon=len(y_test))

3. Predict and evaluate

After fitting, call predict and score against the held-out data.

from yohou.metrics import MeanAbsoluteError
from yohou.plotting import plot_forecast

y_pred = forecaster.predict(forecasting_horizon=len(y_test))
scorer = MeanAbsoluteError()
scorer.fit(y_train)
scorer.score(y_test, y_pred)
plot_forecast(y_test, y_pred, y_train=y_train)

How do I use Yohou?

Full documentation is available at https://yohou.readthedocs.io/.

Interactive examples are available in the examples/ directory:

Can I contribute?

We welcome contributions, feedback, and questions:

If you are interested in becoming a maintainer or taking a more active role, please reach out to the maintainers at .

Where can I learn more?

Here are the main Yohou resources:

For questions and discussions, you can also open a discussion.

License

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

How do I cite Yohou?

If you use Yohou in work you publish, please cite it:

Guillaume Tauzin. Yohou: A time series forecasting package based on Scikit-Learn and Polars. https://github.com/stateful-y/yohou

Or in BibTeX:

@software{yohou,
  author  = "Guillaume Tauzin",
  title   = "{Yohou: A time series forecasting package based on Scikit-Learn and Polars}",
  url     = "https://github.com/stateful-y/yohou",
  license = "Apache-2.0"
}

Reference managers can read CITATION.cff directly. To cite a specific version, see the citation page.

Acknowledgements

We would like to thank Evolta Technologies for their support to the project.


Evolta Technologies


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

About

Time-series forecasting built on scikit-learn, with composable forecasters, transformers and cross-validation.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

7 stars

Watchers

1 watching

Forks

Releases

Contributors

Languages