PKBoost: Adaptive GBDT for Concept Drift, Built from scratch in Rust, PKBoost manages changing data distributions in fraud detection with a fraud rate of 0.2%. It shows less than 2% degradation under drift. In comparison, XGBoost experiences a 31.8% drop and LightGBM a 42.5% drop

No description available

🔍 Enhance your model's performance with PKBoost, an adaptive gradient boosting library that tackles concept drift in imbalanced multi-class data.

Welcome to PkBoost Genesis, an experimental gradient boosting framework blending Newton gain with Shannon entropy-based mutual information, built with a C++ backend. Ideal for education and research, it’s unoptimized . Contributions welcome. MIT-licensed. Contact kharatpushp16@outlook.com for feedback!