Skip to main content
. Author manuscript; available in PMC: 2022 Nov 26.
Published in final edited form as: Physiol Meas. 2021 Nov 26;42(10):10.1088/1361-6579/ac2fb8. doi: 10.1088/1361-6579/ac2fb8

Table 5:

Classifier performance in discriminating TB patients from non-TB patients: The performance achieved by each considered classifier architecture in terms of the area under the ROC curve (AUC). Optimal classifier hyperparameters, determined during cross-validation, are also shown. Mean and standard deviations of AUC are calculated over the five outer folds of the nested k-fold cross-validation, shown in Figure 7.

Classifier Feature Hyperparameters Mean AUC Mean SD Mean Accuracy Mean PPV Mean NPV Optimal Classifier Hyperparameters
LR M=13; F=211;
S=1
0.8000 0.0519 78.82% 75.41% 83.29% ν1=100, ν2=0.15, ν3=0.85
LR M=13; F=210;
S=1
0.7842 0.0645 75.93% 72.59% 80.43% ν1=0.001, ν2=0.25, ν3=0.75
LR M=13; F=211;
S=1
0.7477 0.0402 72.67% 70.36% 75.56% ν1=10, ν2=0.15, ν3=0.85
LR M=26; F=211;
S=1
0.8632 0.0601 84.54% 80.56% 89.71% ν1 =0.01, ν2 =0.3, ν3 =0.7
LR M=26; F=211;
S=1
0.7845 0.0491 76.22% 72.9% 80.67% ν1=0.00001, ν2=0.45, ν3=0.55
LR M=39; F=211;
S=1
0.7458 0.0531 72.11% 70.14% 74.51% ν1=0.0001, ν2=0.7, ν3=0.3
LR M=39; F=211;
S=1
0.7402 0.0455 73.02% 70.57% 76.14% ν1=0.0001, ν2=0.4, ν3=0.6
LR B=60; F=211;
S=1
0.7526 0.0507 72.89% 70.26% 76.31% ν1=0.1, ν2=0.45, ν3=0.55
KNN M=26; F=211;
S=1
0.7701 0.0505 75.09% 71.89% 79.38% χ1=80 χ2=20
KNN M=26; F=211;
S=1
0.7394 0.0385 70.76% 68.95% 72.96% χ1=60 χ2=15
SVM M=26; F=211;
S=1
0.7435 0.0543 71.91% 69.69% 74.7% ζ1=0.01 ζ2=100
SVM M=39; F=210;
S=1
0.7291 0.0495 70.05% 67.95% 72.71% ζ1=0.001 ζ2=0.0001
MLP M=13; F=211;
S=1
0.7389 0.0457 71.16% 68.91% 74.02% ξ1=80, ξ2=0.0001, ξ3=0.65
MLP M=26; F=211;
S=1
0.8000 0.0391 77.87% 75.13% 81.28% ξ1=50, ξ2=0.001, ξ3=0.55
MLP M=39; F=211;
S=1
0.7742 0.0409 76.47% 73.63% 80.09% ξ1=30, ξ2=0.01, ξ3=0.35
CNN M=26; F=211;
S=1
0.7109 0.0409 68.89% 67.7% 70.25% α1=32 α2=0.3, α3=128
CNN M=39; F=210
S=1
0.7001 0.0301 68.71% 67.52% 70.07% α1=64 α2=0.1, α3=128