Table 4.
The results of iERM in the Japanese dataset compared to the traditional classification model.
| Japan | Normal | Stage 1 | Stage 2 | Stage 3 | Stage 4 | Stage 5 |
|---|---|---|---|---|---|---|
| without segmentation | ||||||
| Precision (%) | 98.3 | 33.3 | 5.9 | 71.4 | 0.0 | 80.0 |
| Recall (%) | 96.7 | 40.0 | 33.3 | 54.1 | 0.0 | 57.1 |
| F1-score (%) | 97.5 | 36.4 | 10.0 | 61.5 | 0.0 | 66.7 |
| AUC (%) | 96.5 | 72.3 | 83.3 | 90.9 | 84.5 | 96.0 |
| Accuracy (%) | 73.5 | |||||
| with segmentation | ||||||
| Precision (%) | 89.6 | 33.3 | 25.0 | 82.1 | 100.0 | 100.0 |
| Recall (%) | 98.4 | 60.0 | 66.7 | 62.2 | 25.0 | 57.1 |
| F1-score (%) | 93.7 | 42.9 | 36.6 | 70.8 | 40.0 | 72.7 |
| AUC (%) | 99.9 | 78.7 | 95.9 | 90.2 | 88.5 | 98.1 |
| Accuracy (%) | 79.4 | |||||