minerva-ml/open-solution-home-credit

Implements progressive feature engineering and ensemble stacking across multiple branched solutions, from LightGBM baselines to advanced stacking strategies combining model and feature diversity. Integrates with Neptune.ml for experiment tracking and hyperparameter tuning (random search, choice, uniform, and log-uniform distributions), though runs as standalone Python scripts without required dependencies. Provides six documented solution iterations with cross-validation evaluation and public leaderboard benchmarks, enabling reproducible baseline performance and modular pipeline composition.

460 stars. No commits in the last 6 months.