skforecast
A Python library for time series forecasting using scikit-learn compatible machine learning models.
What is skforecast?
Skforecast is a Python library for time series forecasting using machine learning models. It provides a scikit-learn compatible API that allows data scientists and developers to build, validate, and deploy forecasting models with popular estimators like LightGBM, XGBoost, and CatBoost. The library simplifies the entire forecasting workflow, from feature engineering to production deployment.
Data scientists, machine learning engineers, and analysts who need to build and deploy time series forecasting models using Python and scikit-learn compatible tools.
Developers choose Skforecast for its seamless integration with the scikit-learn ecosystem, comprehensive forecasting capabilities, and production-ready tooling that simplifies the entire model development lifecycle.
Overview
Python library for time series forecasting using scikit-learn compatible models, statistical methods, and foundation models
Use Cases
Best For
- Building time series forecasting models with scikit-learn compatible estimators
- Multi-series and multivariate forecasting projects
- Production deployment of machine learning forecasting pipelines
- Hyperparameter tuning and backtesting for time series models
- Educational purposes and learning time series forecasting with Python
- Rapid prototyping of forecasting solutions with flexible strategies
Not Ideal For
- Real-time streaming forecasting applications requiring immediate model updates and low-latency predictions
- Projects deeply integrated with pure deep learning frameworks like PyTorch without scikit-learn wrappers
- Use cases needing extensive built-in visualization tools for exploratory time series analysis
- Situations where simple statistical methods (e.g., exponential smoothing) suffice and machine learning adds unnecessary complexity
Pros & Cons
Pros
Seamlessly works with any scikit-learn compatible estimator, including LightGBM, XGBoost, and CatBoost, as demonstrated in the quick example using LGBMRegressor.
Offers multiple forecaster types like Recursive, Direct, and MultiSeries, catering to different needs such as single or multi-series forecasting, shown in the forecaster table.
Includes comprehensive methods for backtesting, hyperparameter tuning, and model validation, ensuring robust deployment, highlighted in the production-ready tools section.
Enables uncertainty quantification across several forecaster types, which is crucial for risk-aware decision making in predictions.
Provides machine-readable context files for LLMs to generate accurate code, speeding up development, as mentioned in the AI-assisted forecasting section.
Cons
Restricted to estimators compatible with scikit-learn, so native deep learning models (e.g., pure TensorFlow) require wrappers, limiting direct integration.
With eight different forecaster classes, users might struggle to choose the right one, adding a learning curve despite the unified API.
Advanced features like RNN support or full probabilistic forecasting may need extra libraries (e.g., TensorFlow), increasing setup overhead beyond the basic install.
While it includes statistical forecasters like ForecasterStats, the library is primarily ML-oriented, which might not suit projects preferring classic time series models without ML integration.
Frequently Asked Questions
Related Projects
PyTorch - Tensors and Dynamic neural networks in Python with strong GPU accelerationTensors and Dynamic neural networks in Python with strong GPU acceleration
kerasDeep Learning for humans
streamlitStreamlit — A faster way to build and share data apps.
gradioBuild and share delightful machine learning apps, all in Python. 🌟 Star to support our work!
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.