Distributional Gradient Boosting Machines

TL;DR

This paper introduces a unified probabilistic gradient boosting framework that models the entire conditional distribution of a response variable, enabling probabilistic forecasting and improved accuracy in regression tasks.

Contribution

It presents a novel framework combining distributional modeling with gradient boosting, utilizing likelihood-based methods and Normalizing Flows for flexible distributional predictions.

Findings

Achieves state-of-the-art forecast accuracy

Enables creation of prediction intervals and quantiles

Integrates with XGBoost and LightGBM for practical use

Abstract

We present a unified probabilistic gradient boosting framework for regression tasks that models and predicts the entire conditional distribution of a univariate response variable as a function of covariates. Our likelihood-based approach allows us to either model all conditional moments of a parametric distribution, or to approximate the conditional cumulative distribution function via Normalizing Flows. As underlying computational backbones, our framework is based on XGBoost and LightGBM. Modelling and predicting the entire conditional distribution greatly enhances existing tree-based gradient boosting implementations, as it allows to create probabilistic forecasts from which prediction intervals and quantiles of interest can be derived. Empirical results show that our framework achieves state-of-the-art forecast accuracy.