Table 20 Cross-domain adaptation results.
From: PatternFusion: a hybrid model for pattern recognition in time-series data using ensemble learning
Model configuration | Target domain | Accuracy (%) | F1-score (%) | AUC (%) | EER (%) |
|---|---|---|---|---|---|
PatternFusion (baseline) | PolyU | 95.8 | 93.7 | 97.6 | 1.5 |
BioCop | 94.7 | 92.5 | 96.3 | 1.8 | |
Cross-domain transfer set | 89.4 | 87.1 | 91.2 | 3.5 | |
PatternFusion + fine-tuning | PolyU | 96.3 | 94.2 | 98.1 | 1.3 |
BioCop | 95.1 | 93.1 | 97.0 | 1.6 | |
Cross-Domain Transfer Set | 92.7 | 90.8 | 94.5 | 2.8 | |
PatternFusion + domain-adversarial training | PolyU | 96.7 | 94.5 | 98.4 | 1.1 |
BioCop | 95.6 | 93.8 | 97.5 | 1.4 | |
Cross-Domain Transfer Set | 93.5 | 91.4 | 95.2 | 2.5 |
