BoXHED2.0

Boosted eXact Hazard Estimator with Dynamic covariates v2.0 (BoXHED2.0, pronounced 'box-head') is a software package designed for nonparametric estimation of hazard functions using gradient-boosted trees. BoXHED2.0 supports both time-static and time-dependent covariates.

This document provides instructions for manually installing the BoXHED2.0 package. For additional information, including precompiled packages, tutorials, and related publications, please visit the package's GitHub page.

Prerequisites

BoXHED2.0 has been developed and tested on Linux and macOS. The following are required:

• cmake (>=3.18.2) (Note for Windows users: CMake is included in the Visual Studio installation.)
• Python (=3.8)
• CUDA (=11.1) (Required only if GPU support is needed.)
• conda (we recommend using the free Anaconda distribution)

For Windows users, the Visual Studio 17 2022 toolset must be installed. During installation, select "Desktop Development with C++" under the "Workloads" tab in the "Desktop and Mobile" section. Ensure the following options are selected:

Installing BoXHED2.0 (multicore CPU support)

Installing BoXHED2.0 (multicore CPU + GPU support)

Installing SHAP for BoXHED