EpistasisLab/tpot

Employs a graph-based pipeline representation with multi-objective optimization and genetic feature selection, enabling discovery of both preprocessing and modeling stages simultaneously. Integrates with scikit-learn, XGBoost, and LightGBM while leveraging Dask for distributed parallel evolution across multiple processes. Recent refactoring introduced modular architecture for customizable evolutionary algorithms and expanded search space definitions beyond traditional tree structures.

10,049 stars. Used by 1 other package. No commits in the last 6 months. Available on PyPI.