Skip to main content
. 2022 Jan 3;8:781937. doi: 10.3389/fmed.2021.781937

Table 3.

The performance of the four classifier models.

Fold1 Fold2 Fold3 Fold4 Fold5 Mean-Value
(A) GBDT-LZC
Accuracy (%) 67 87 80 93 79 81
Precision (%) 69 80 81 100 90 84
Recall (%) 90 100 90 92 82 91
F1-score (%) 78 89 86 96 86 87
AUC (%) 64 89 82 100 70 81
Sensitivity (%) 90 100 90 92 82 91
Specificity (%) 20 71 60 100 67 64
(B) SVM-LZC
Accuracy (%) 60 53 67 67 64 62
Precision (%) 70 56 78 100 80 77
Recall (%) 70 63 70 62 73 67
F1-score (%) 70 59 74 76 76 71
AUC (%) 54 64 64 100 33 63
Sensitivity (%) 70 63 70 62 73 67
Specificity (%) 40 43 60 100 33 55
(C) GBDT-KC
Accuracy (%) 67 87 80 100 79 82
Precision (%) 69 80 82 100 90 84
Recall (%) 90 100 90 100 82 92
F1-score (%) 78 89 86 100 86 88
AUC (%) 66 89 88 100 73 83
Sensitivity (%) 66 89 88 100 73 83
Specificity (%) 90 100 90 100 82 92
(D) SVM-KC
Accuracy (%) 60 53 67 67 64 62
Precision (%) 70 56 78 100 80 77
Recall (%) 70 63 70 62 73 67
F1-score (%) 70 59 74 76 76 71
AUC (%) 54 63 64 100 33 63
Sensitivity (%) 70 63 70 62 73 67
Specificity (%) 40 43 60 100 33 55

LZC, Lempel-Ziv complexity; GBDT, gradient boosting decision tree; AUC, the area under the curve; SVM, support vector machine; RFE, recursive feature elimination; KC, Kolmogorov complexity.