Table 5 Ablative analysis of each component in the proposed neural architecture.
From: Dual graph attention network for robust fault diagnosis in photovoltaic inverters
Components | F1-score | Test acc | Precision | Recall |
|---|---|---|---|---|
DGAT (✗ FeatureGAT) | 0.812 | 91.27% | 0.823 | 0.801 |
DGAT (✗ TempGAT) | 0.781 | 87.62% | 0.794 | 0.769 |
DGAT (✗ Regularizer) | 0.835 | 90.13% | 0.846 | 0.819 |
DGAT (✗ Cross Attention) | 0.882 | 92.51% | 0.903 | 0.874 |
DGAT (✓ All Components) | 0.971 | 97.35% | 0.979 | 0.982 |
