overhang_surrogates: A Python package for sampling, training and visualising surrogate models for building energy simulations

Abstract

We present overhang_surrogates, a lightweight Python package that streamlines surrogate-model workflows for building-energy studies. It provides space-filling Monte Carlo sampling utilities including a Python reimplementation of the MIPT sampler with incremental extension, helpers to build batched building energy model samples and run EnergyPlus simulations, a simple interface for k[jls-end-space/]-fold cross-validated XGBoost ensembles and grid predictions, and a vedo-based 3D plotting helper. By automating sampling, batched simulation, ensemble training, prediction and visualisation, the package shortens time-to-prototype and lowers the barrier to reproduce and extend simulation driven surrogate experiments. The software is open-source and designed for easy reuse and extension.