Skip to main content

Table 3 Performance comparison of different models to predict 2-year relapse in the validation set

From: Multi-institutional development and external validation of machine learning-based models to predict relapse risk of pancreatic ductal adenocarcinoma after radical resection

Model AUC 95%CI.lower 95%CI.upper Sensitivity Specificity Accuracy PPV NPV F1 RMSE
LR 0.625 0.482 0.760 0.847 0.350 0.722 0.794 0.438 0.820 0.467
RF 0.655 0.518 0.784 0.898 0.250 0.734 0.779 0.455 0.835 0.431
SVM 0.597 0.460 0.731 0.831 0.200 0.671 0.754 0.286 0.790 0.471
GBM 0.652 0.517 0.776 0.831 0.250 0.684 0.766 0.333 0.797 0.463
NN 0.608 0.464 0.747 0.831 0.200 0.671 0.754 0.286 0.790 0.450
KNN 0.689 0.558 0.817 0.915 0.200 0.734 0.771 0.444 0.837 0.416