Table 4.

Comparison results for the electrocardiogram dataset.^a

Experiments	AUROCb (95% CI)	F1-score (95% CI)	Precision (95% CI)	Recall (95% CI)
Benchmark	0.954 (0.930-0.978)	0.814 (0.655-0.910)	0.820 (0.672-0.943)	0.814 (0.640-0.936)
Basic FLc	0.938 (0.860-0.978)	0.807 (0.651-0.931)	0.823 (0.645-0.942)	0.795 (0.660-0.925)
Imbalanced FL	0.943 (0.883-0.977)	0.807 (0.635-0.902)	0.830 (0.650-0.935)	0.788 (0.626-0.905)

^aAll results are presented with a 95% CI by resampling 100 times.

^bAUROC: area under the receiver operating characteristic curve.

^cFL: federated learning.