Skip to main content
. 2017 Nov 7;7:14738. doi: 10.1038/s41598-017-15137-7

Table 3.

Model performance in training and testing data for the best selected classification models for each dataset. PPV = Positive predictive value. NPV = Negative predictive value. CI = confidence interval (95%).

Model Method Train accuracy Train Kappa Test Accuracy Test Kappa Test Sensitivity Test Specificity Test PPV Test NPV Tuning parameter Rank
E-Clinical (N = 284) First Model Random Forest 0.608[95% CI = 0.598–0.618] 0.052 [95% CI = 0.028–0.075] 0.586 −0.022 0.61 0.33 0.91 0.07 Mtry = 2 1
E-Clinical (N = 284) Second Model Generalised Linear model 0.574 [95% CI = 0.561–0.587] 0.056[95% CI = 0.028–0.083] 0.600 0.132 0.66 0.48 0.72 0.41 Alpha = 0.1, lambda = 0.019 (1)
A-Clinical, genetic, Expression (N = 108) K-Nearest Neighbours (KNN) 0.591 [95% CI = 0.556–0.625] 0.096 [95% CI = 0.022–0.170] 0.577 0.077 0.60 0.50 0.80 0.27 K = 9  = 2
B-Clinical, genetic (N = 108) KNN 0.591 [95% CI = 0.556–0.625] 0.096 [95% CI = 0.022–0.170] 0.577 0.077 0.60 0.50 0.80 0.27 K = 9  = 2
D-Clinical, Expression (N = 108) KNN 0.591 [95% CI = 0.556–0.625] 0.096 [95% CI = 0.022–0.170] 0.577 0.077 0.60 0.50 0.80 0.27 K = 9  = 2
C-Clinical N = 108 KNN 0.591 [95% CI = 0.556–0.625] 0.096 [95% CI = 0.022–0.170] 0.577 0.077 0.60 0.50 0.80 0.27 K = 9  = 2