
Website: auto.gluon.ai
Brief Description: AutoGluon is an open source AutoML framework that focuses on accuracy, speed, and ease of use. With just three lines of code, users can train accurate ML models for tabular, time series, or multimodal data. AutoGluon automatically builds a diverse set of models and ensembles them to maximize the predictive accuracy. The library is built for both beginners, who want simple and effective automation, and advanced users, who need flexibility to customize and extend their models.
Key Features with their Benefits:
- Multi-layer Stacking & Ensembling: Uses advanced ensembling techniques like stacking and weighted averaging to maximize predictive accuracy.
- Model Portfolios: AutoGluon trains a diverse collection of models, allowing it to automatically select and ensemble the best ones for the given ML task.
- Automated Data Cleaning: Automatically deals with preprocessing issues such as missing values handling or categorical encoding, reducing the need for manual data preparation.
- Scalability: Designed to be efficient and easy to run, whether on a laptop or in a production environment in the cloud.
Keywords: AutoML, Open Source, Python, Ensembling, Stacking, Tabular, Multimodal, Time Series
Installation:
`pip install autogluon`
For more details, see the installation instructions.
Exemplary Usage:
- Tabular classification & regression
- Time series forecasting
- ML for text, images and multimodal data
- AutoGluon competition winning solutions
How to cite: See CITING.md.
Contributed by Nick Erickson and Oleksandr Shchuhr (AutoGluon Team; January 2025)