Table 5.

PAF records classification results for different values of N and m. These results should be compared with those in Table 4, where the same dataset was used, but the complete time series instead. For lengths around 50 samples, classification performance is very similar to that achieved with the entire records.

m	N	Sensitivity	Specificity	p	AUC
3	10	0.52	0.68	1.0000	0.5000
3	25	0.68	0.56	0.0857	0.6416
3	40	0.68	0.72	0.0045	0.7336
3	45	0.76	0.84	0.0002	0.8048
3	50	0.80	0.80	0.0002	0.7984
3	60	0.84	0.72	0.0003	0.7920
3	75	0.76	0.76	0.0004	0.7904
3	100	0.92	0.60	0.0003	0.7920
4	10	0.64	0.52	0.1278	0.6184
4	25	0.52	0.68	0.2169	0.6016
4	50	0.72	0.76	0.0004	0.7904
4	75	0.80	0.72	0.0003	0.7936
4	100	0.88	0.68	0.0001	0.8096
4	150	0.92	0.68	<0.0001	0.8496
5	10	0.00	1.00	0.8083	0.5200
5	25	0.52	0.60	0.2192	0.5984
5	50	0.68	0.84	0.0012	0.7664
5	75	0.60	0.84	0.0007	0.7784
5	100	0.76	0.72	0.0017	0.7584
5	200	0.88	0.64	0.0001	0.8208