Skip to main content

Table 2 Performance of risk prediction models in the test cohort

From: Comparing the performance of machine learning and conventional models for predicting atherosclerotic cardiovascular disease in a general Chinese population

Algorithms

Overall

Discrimination

Calibration

Clinical Usefulness

Brier

MCC

AUC (95%CI)

Hosmer–Lemeshow χ2 ( p value)

Net benefit at threshold of 5%

PCE (White)

0.045

0.186

0.777 (0.733–0.821)

37.3 (p < 0.01)

0.013

China-PAR

0.043

0.191

0.780 (0.737–0.822)

67.6 (p < 0.001)

0.016

RePCE (White)

0.057

0.194

0.779 (0.734–0.825)

126.6 (p < 0.001)

0.016

ReChina-PAR

0.043

0.193

0.780 (0.737–0.822)

18.6 (p < 0.05)

0.017

ANN

0.041

0.218

0.800 (0.759–0.838)

9.1 (p = 0.33)

0.017

RF

0.042

0.181

0.759 (0.713–0.804)

12.6 (p = 0.13)

0.011

GBM

0.042

0.193

0.774 (0.727–0.820)

12.1 (p = 0.15)

0.013

KNN

0.042

0.175

0.767 (0.723–0.811)

34.7 (p < 0.01)

0.015

Adaboost

0.051

0.163

0.727 (0.679–0.775)

136.4 (p < 0.001)

0.010

SVM

0.043

0.145

0.697 (0.642–0.752)

4.0 (p = 0.86)

0.009

Catboost

0.041

0.206

0.787 (0.745–0.830)

10.2 (p = 0.25)

0.015

  1. Abbreviations: ASCVD atherosclerotic cardiovascular disease, CI confidence interval, MCC Matthews correlation coefficient, AUC area under the receiver operating characteristic curve, PCE Pooled Cohort Equations, China-PAR Prediction for ASCVD Risk in China, RePCE Recalibrated PCE, ReChina-PAR Recalibrated China-PAR, ANN Artificial Neural Network, RF Random Forest, GBM Gradient Boosting Machine, KNN K Nearest Neighbor, Adaboost Adaptive Boosting, SVM Support Vector Machine, Catboost Categorical Boosting
Back to article page

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us