Table 4 The second dataset’s (Stratified K Fold split) experimental results include a classification report from (a) standard threshold \(\lambda = 0.5\), and (b) with threshold moving optimization \(\lambda = 0.44\).
From: Enhancing cardiotocography classification via ensemble learning and threshold optimization
Precision | Recall | F1-score | Support | Precision | Recall | F1-score | Support | ||
|---|---|---|---|---|---|---|---|---|---|
No threshold | Threshold moving | ||||||||
Normal | 1.00 | 1.00 | 1.00 | 117 | Normal | 1.00 | 1.00 | 1.00 | 117 |
Suspicious | 0.76 | 0.50 | 0.60 | 38 | Suspicious | 0.79 | 0.29 | 0.42 | 38 |
Pathological | 0.24 | 0.50 | 0.32 | 12 | Pathological | 0.25 | 0.75 | 0.38 | 12 |
Accuracy | – | – | 0.85 | 167 | Accuracy | – | – | 0.82 | 167 |
Macro avg | 0.67 | 0.67 | 0.64 | 167 | Macro avg | 0.68 | 0.68 | 0.60 | 167 |
Weighted avg | 0.89 | 0.85 | 0.86 | 167 | Weighted avg | 0.90 | 0.82 | 0.82 | 167 |
