ngboost

Natural Gradient Boosting for Probabilistic Prediction

Evolutionary Neural Architecture Search on Transformers for RUL Prediction

Play around with NGBoost and compare with LightGBM and XGBoost

Probabilistic Time Series Forecasting

ML project that predicts wars with 95% classification accuracy

An R interface to the NGBoost.

R wrapper for NGBoost python module

A CLI Tuner of NGBoost

Bachelor's thesis: benchmarking prediction intervals (LRM, quantile regression, gradient boosting, NGBoost) with split conformal calibration in a Monte Carlo study — reproducible notebook + Docker.

Bit-identical NGBoost prediction, 3.5-3.7x faster one-row by checking the input once per call, plus a fit-time warning when every tree is constant (the only case where folding speeds it up)

ML pipeline for bank insolvency prediction (1–5 years). 4 ensemble models + SHAP explainability. Production-ready with Docker, reproducible research.

Probabilistic Hong Kong horse-racing predictions with NGBoost and Monte Carlo simulation

RUL prediction for turbofan engines on NASA N-CMAPSS (DS02): classical ML (NGBoost + conformal prediction) vs. deep learning (CNN deep ensemble) with calibrated uncertainty quantification.

Risk aware battery sizing for a PV and storage EV charging hub in Bielefeld

Forecasts seven NBA player box-score stats as full probability distributions, with the data pipeline, leakage checks and evaluation behind them. Personal project.

Replication of NGBoost (Duan et al., ICML 2020) . Runs the released code on Yacht and Kin8nm, tests its stage-selection rule, and extends the natural, ordinary and constant-scale comparison to five UCI datasets and synthetic data with known noise. Includes the report, scripts, prepared data, saved runs and an LLM-as-judge evaluation.