bankruptcy-prediction: v1.0.0 - Bankrupty Prediction

🚀 v1.0.0 - Initial Bankrupty Prediction Pipeline

This version transitions from Jupyter notebooks to production CLI scripts with enterprise-grade reproducibility, logging, and explainability for financial risk assessment.

Highlights:

1. Unified Training Framework: Single training_utils.py eliminates code duplication across 3 CLI entry points
2. 4 Ensemble Models × 5 Horizons: NGBoost, Random Forest, XGBoost, LightGBM trained sequentially across 1–5 year prediction horizons
3. Reproducible Hyperparameters: Seeded RNG + hardcoded 0.4 threshold; XGBoost saves 100 iterations to CSV + consolidated best_params.json
4. SHAP Explainability: Unified analyzer supporting all model types (TreeExplainer for tree models, KernelExplainer for NGBoost)
5. Production-Ready: Time-based splits (prevent temporal leakage), class imbalance handling, standardized CSV output format with risk buckets + confusion types