Skip to main content

Table 2 Additional test performance metrics

From: Predicting two-year survival versus non-survival after first myocardial infarction using machine learning and Swedish national register data

Model

Sens/Spec

PPV/NPV

Detection rate

Detection incidence

Accuracy (95% CI)

Full predictor set (n = 39)

 LR

0.771/0.770

0.293/0.965

0.085

0.290

0.770 (0.764 to 0.776)

 C5.0

0.798/0.739

0.275/0.967

0.088

0.320

0.746 (0.740 to 0.752)

 RF

0.789/0.752

0.282/0.966

0.087

0.307

0.756 (0.750 to 0.762)

 SVM

0.784/0.751

0.280/0.966

0.086

0.308

0.755 (0.749 to 0.761)

Reduced predictor set (n = 10)

 LR

0.754/0.758

0.278/0.961

0.083

0.298

0.758 (0.752 to 0.763)

 C5.0

0.768/0.757

0.281/0.964

0.084

0.301

0.758 (0.752 to 0.764)

 RF

0.771/0.746

0.272/0.963

0.085

0.311

0.748 (0.742 to 0.754)

 SVM

0.751/0.756

0.275/0.961

0.083

0.300

0.755 (0.749 to 0.761)

Minimal predictor set (n = 5)

 LR

0.749/0.750

0.270/0.960

0.082

0.305

0.750 (0.744 to 0.756)

 C5.0

0.758/0.736

0.262/0.961

0.083

0.319

0.738 (0.732 to 0.744)

 RF

0.755/0.703

0.239/0.959

0.083

0.348

0.708 (0.702 to 0.715)

 SVM

0.732/0.753

0.268/0.958

0.080

0.300

0.751 (0.745 to 0.757)

  1. Results of trained models on 100% of testing data (n = 20,777) by predictor set. For all models, Base Rate Incidence = 0.110, and No Information Rate = 0.890
  2. Sens sensitivity, Spec specificity, PPV positive predictive value, NPV negative predictive value, CI confidence interval, NIR no information rate, C5.0 C5.0 boosted decision trees, LR logistic regression, RF random forest, SVM support vector machine