GPBoost

Why we included this project

GPBoost is worth a look when your tabular data has structure that plain tree ensembles miss, like spatial correlation, grouped observations, or repeated measurements on the same subjects. It combines gradient-boosted trees with Gaussian process and random-effects terms, so a single fitted model can capture a non-linear fixed-effects function while still accounting for clustering or smooth spatial variation. That makes it a natural fit for longitudinal and panel data, high-cardinality categorical variables, and spatial or spatio-temporal prediction where you want continuous smoothness across space. The C++ core ships with both Python and R packages, and the documentation includes worked examples for these scenarios, so it's practical to evaluate against your own datasets. If you've been torn between a boosting library and a mixed-effects or GP model, this lets you do both in one place.