skforecast 0.8.1

Forecasting time series with scikit-learn regressors. It also works with any regressor compatible with the scikit-learn API (pipelines, CatBoost, LightGBM, XGBoost, Ranger...).

Skforecast is a Python library that eases using scikit-learn regressors as single and multi-step forecasters. It also works with any regressor compatible with the scikit-learn API (LightGBM, XGBoost, CatBoost, ...).

Why use skforecast?

The fields of statistics and machine learning have developed many excellent regression algorithms that can be useful for forecasting, but applying them effectively to time series analysis can still be a challenge. To address this issue, the skforecast library provides a comprehensive set of tools for training, validation and prediction in a variety of scenarios commonly encountered when working with time series. The library is built using the widely used scikit-learn API, making it easy to integrate into existing workflows. With skforecast, users have access to a wide range of functionalities such as feature engineering, model selection, hyperparameter tuning and many others. This allows users to focus on the essential aspects of their projects and leave the intricacies of time series analysis to skforecast. In addition, skforecast is developed according to the following priorities:

• Fast and robust prototyping. :zap:
• Validation and backtesting methods to have a realistic assessment of model performance. :mag:
• Models must be deployed in production. :hammer:
• Models must be interpretable. :crystal_ball:

Documentation: https://skforecast.org :books:

Installation

The default installation of skforecast only installs hard dependencies.

pip install skforecast

Specific version:

pip install skforecast==0.8.1

Latest (unstable):

pip install git+https://github.com/JoaquinAmatRodrigo/skforecast#master

Install the full version (all dependencies):

pip install skforecast[full]

Install optional dependencies:

pip install skforecast[sarimax]

pip install skforecast[plotting]

Dependencies

• Python >= 3.8

Hard dependencies

• numpy>=1.20, <1.25
• pandas>=1.2, <2.1
• tqdm>=4.57.0, <4.65
• scikit-learn>=1.0, <1.3
• optuna>=2.10.0, <3.2
• joblib>=1.1.0, <1.3.0

Optional dependencies

• matplotlib>=3.3, <3.8
• seaborn>=0.11, <0.13
• statsmodels>=0.12, <0.14
• pmdarima>=2.0, <2.1

Features

• Create recursive autoregressive forecasters from any regressor that follows the scikit-learn API
• Create direct autoregressive forecasters from any regressor that follows the scikit-learn API
• Create multi-series autoregressive forecasters from any regressor that follows the scikit-learn API
• Include exogenous variables as predictors
• Include custom predictors (rolling mean, rolling variance ...)
• Multiple backtesting methods for model validation
• Grid search, random search and Bayesian search to find optimal lags (predictors) and best hyperparameters
• Include custom metrics for model validation and grid search
• Prediction interval estimated by bootstrapping and quantile regression
• Get predictor importance
• Forecaster in production

What is new in skforecast 0.8.1?

• Support for pandas 2.0.x.
• New user guide on how to include categorical variables in the Forecasters.
• New user guide on how to use GPU in Google Colab with XGBoost and LightGBM regressors.
• Include custom kwargs during fit.
• The dtypes of exogenous variables are maintained when generating the training matrices with the create_train_X_y method in all the Forecasters.
• Include gap argument in backtesting functions to omit observations between training and prediction.
• Bug fixes and performance improvements.

Visit the release notes to view all notable changes.

Documentation

The documentation for the latest release is at skforecast docs.

Recent improvements are highlighted in the release notes.

• Introduction to time series and forecasting

• Recursive multi-step forecasting

• Independent multi-series forecasting

• Dependent multi-series forecasting (Multivariate forecasting)

• Backtesting (validation) of forecasting models

• Hyperparameter tuning and lags selection of forecasting models

• Probabilistic forecasting

• Using forecasters in production

Examples and tutorials

English

• Skforecast: time series forecasting with Python and Scikit-learn

• Forecasting electricity demand with Python

• Forecasting web traffic with machine learning and Python

• Forecasting time series with gradient boosting: skforecast, XGBoost, LightGBM and CatBoost

• Bitcoin price prediction with Python

• Prediction intervals in forecasting models

• Multi-series forecasting

• Reducing the influence of Covid-19 on time series forecasting models

• Forecasting time series with missing values

• Intermittent demand forecasting

Español

• Skforecast: forecasting series temporales con Python y Scikit-learn

• Forecasting de la demanda eléctrica

• Forecasting de las visitas a una página web

• Forecasting series temporales con gradient boosting: skforecast, XGBoost, LightGBM y CatBoost

• Predicción del precio de Bitcoin con Python

• Workshop predicción de series temporales con machine learning Universidad de Deusto / Deustuko Unibertsitatea

• Intervalos de predicción en modelos de forecasting

• Multi-series forecasting

• Predicción de demanda intermitente

Donating

If you found skforecast useful, you can support us with a donation. Your contribution will help to continue developing and improving this project. Many thanks!

How to contribute

For more information on how to contribute to skforecast, see our Contribution Guide.

Citation

If you use this software, please cite it using the following metadata.

APA:

Amat Rodrigo, J., & Escobar Ortiz, J. skforecast (Version 0.8.1) [Computer software]

BibTeX:

@software{skforecast,
author = {Amat Rodrigo, Joaquin and Escobar Ortiz, Javier},
license = {MIT},
month = {5},
title = {{skforecast}},
version = {0.8.1},
year = {2023}
}

View the citation file.

License

BSD 3-Clause License