Research of GPCEMBP Glucose Prediction Algorithm Based on Continuous Glucose Monitoring

Table 3

Selection of the number of inputs and the BP network architectures.

pH (min)	Inputs	Network architecture	Output	FIT (%)	RMSE (mg/dl)	NPE (%)	ET (s)
63	7	15-13-1	1	91.14	16.23	7.79	6.63
63	12	15-13-1	1	95.66	11.51	5.52	16.78
63	15	15-13-1	1	97.25	9.17	4.40	5.34
63	20	15-13-1	1	97.87	8.1	3.89	11.48
63	25	15-13-1	1	97.94	7.97	3.83	10.9
63	12	13-12-1	1	92.11	15.36	7.37	6.77
63	12	17-14-1	1	95.34	11.9	5.71	4.19
63	12	19-15-1	1	94.91	12.43	5.97	8.34
63	12	21-16-1	1	96.23	10.72	5.15	22.55
63	12	23-17-1	1	93.56	13.95	6.70	10.07
63	15	13-12-1	1	92.95	14.55	6.99	19.1
63	15	17-14-1	1	95.86	11.23	5.39	11.99
63	15	19-15-1	1	95.88	11.2	5.38	25.61
63	15	21-16-1	1	97.69	8.44	4.05	26.83
63	15	23-17-1	1	97.99	7.9	3.80	24.57