Table 2 Predictive performance of arterial catheterization according to each modeling method using the deep or machine learning technique and a combination of features.

From: Preoperative prediction of the need for arterial and central venous catheterization using machine learning techniques

Features

Model

AUROC

AUPRC

F1-score

Preoperative data

DNN

0.7835 ± 0.0016

0.5296 ± 0.0020

0.3939 ± 0.0233

XGBoost

0.6017 ± 0.0008

0.3385 ± 0.0014

0.3542 ± 0.0019

RF

0.5947 ± 0.0012

0.3302 ± 0.0015

0.3368 ± 0.0031

LR

0.5985 ± 0.0008

0.3358 ± 0.0012

0.3464 ± 0.0018

Laboratory data

DNN

0.6050 ± 0.0107

0.3061 ± 0.0093

0.0865 ± 0.0311

XGBoost

0.5208 ± 0.0005

0.2555 ± 0.0010

0.0960 ± 0.0014

RF

0.5196 ± 0.0010

0.2500 ± 0.0011

0.1106 ± 0.0037

LR

0.5008 ± 0.0002

0.2367 ± 0.0008

0.0094 ± 0.0006

Operation code

DNN

0.8930 ± 0.0007

0.7548 ± 0.0015

0.6770 ± 0.0021

XGBoost

0.6765 ± 0.0008

0.4641 ± 0.0017

0.5188 ± 0.0018

RF

0.5288 ± 0.0049

0.2760 ± 0.0067

0.1095 ± 0.0175

LR

0.7338 ± 0.0009

0.5293 ± 0.0017

0.6226 ± 0.0016

All features

DNN

0.9089 ± 0.0093

0.7943 ± 0.0118

0.4352 ± 0.0760

XGBoost

0.7262 ± 0.0010

0.5292 ± 0.0017

0.6121 ± 0.0018

RF

0.5444 ± 0.0050

0.2952 ± 0.0064

0.1659 ± 0.0169

LR

0.6244 ± 0.0009

0.3547 ± 0.0014

0.4105 ± 0.0018

  1. Data represent means (95% confidence intervals).
  2. AUROC, area under receiver operating characteristic; AUPRC, area under precision-recall curve; DNN, deep neural network; XGBoost, extreme gradient boosting; DT, decision tree; RF, random forest; LR, logistic regression; ASA-PS, American Society of Anesthesiologists physical status.
Back to article page