MURAL: An Unsupervised Random Forest-based Embedding for Electronic Health Record Data

Constructing Unsupervised Random Forests with Missing Values in Clinical Data with Meaningful Missingness

Introduction

MURAL is a Python package for constructing random forests in an unsupervised manner from data with variables that have missing values. The motivation behind the algorithmic design was to model clinical data (specifically, electronic health record data), with variables that have values missing not at random.

The goal of MURAL is to construct a meaningful representation of data with missingness in a manifold forest that can be visualized using manifold based methods, such as the Potential of Heat-diffusion for Affinity-based Trajectory Embedding (PHATE).