Skip to main content

Table 4 Performance of the three ML algorithms using the 16-variable dataset with bootstrapping and of the HellenicSCORE, against the 10-year CVD (fatal or non-fatal) incidence of ATTICA study participants

From: Machine learning methodologies versus cardiovascular risk scores, in predicting disease risk

Algorithm Accuracy Specificity Sensitivity PPV NPV
k-NN 0.65 0.56 0.67 0.89 0.24
RF 0.83 0.46 0.89 0.90 0.45
DT 0.80 0.53 0.85 0.91 0.40
HellenicSCORE 0.85 0.20 0.97 0.87 0.58