Multi-objective Symbolic Regression to Generate Data-driven, Non-fixed Structure and Intelligible Mortality Predictors using EHR: Binary Classification Methodology and Comparison with State-of-the-art

TY - JOUR

T1 - Multi-objective Symbolic Regression to Generate Data-driven, Non-fixed Structure and Intelligible Mortality Predictors using EHR

T2 - Binary Classification Methodology and Comparison with State-of-the-art

AU - Ferrari, Davide

AU - Guidetti, Veronica

AU - Wang, Yanzhong

AU - Curcin, Vasa

N1 - ©2022 AMIA - All rights reserved.

PY - 2023/4/29

Y1 - 2023/4/29

N2 - Symbolic Regression (SR) is a data-driven methodology based on Genetic Programming, and it is widely used to produce arithmetic expressions for modelling learning tasks. Compared to other popular statistical techniques, SR outcomes are given by an arbitrary set of mathematical operations, representing arbitrarily complex linear and non-linear functions without a predefined fixed structure. Another advantage is that, unlike other machine learning algorithms, SR produces interpretable results. In this paper, we explore the qualities and limitations of this technique in a novel implementation as a binary classifier for in-hospital or short-term mortality prediction in patients with Covid-19. Our results highlight that SR provides a competitive alternative to popular statistical and machine learning methodologies to model relevant clinical phenomena thanks to good classification performance, stability in unbalanced dataset management, and intrinsic interpretability.

AB - Symbolic Regression (SR) is a data-driven methodology based on Genetic Programming, and it is widely used to produce arithmetic expressions for modelling learning tasks. Compared to other popular statistical techniques, SR outcomes are given by an arbitrary set of mathematical operations, representing arbitrarily complex linear and non-linear functions without a predefined fixed structure. Another advantage is that, unlike other machine learning algorithms, SR produces interpretable results. In this paper, we explore the qualities and limitations of this technique in a novel implementation as a binary classifier for in-hospital or short-term mortality prediction in patients with Covid-19. Our results highlight that SR provides a competitive alternative to popular statistical and machine learning methodologies to model relevant clinical phenomena thanks to good classification performance, stability in unbalanced dataset management, and intrinsic interpretability.

KW - Humans

KW - COVID-19

KW - Algorithms

KW - Machine Learning

M3 - Article

C2 - 37128446

SN - 1559-4076

VL - 2022

SP - 442

EP - 451

JO - Proceedings / AMIA

JF - Proceedings / AMIA

IS - eCollection

ER -