Regularised Model Identification Improves Accuracy of Multisensor Systems for Noninvasive Continuous Glucose Monitoring in Diabetes Management

Table 1

Model test performance when “part 1” of the data set is used for model identification and “part 2” for model test. In brackets is the complexity model parameter chosen by means of cross-validation. Mean and standard deviation (in brackets) over the experimental sessions for root mean squared error (RMSE), mean absolute difference (MAD), mean absolute relative difference (MARD), error grid analysis (EGA (Clarke)) (A + B (A) C/D/E regions whose sum accounts for 100% of data points), continuous error grid analysis (CEGA) (A_R + B_R (A_R) C_R/D_R/E_R regions whose sum accounts for 100% of data points).

	RMSE (mg/dL)	MAD (mg/dL)	MARD (%)	EGA; A + B (A) C/D/E	CEGA; AR+ BR (AR) CR/DR/ER
LASSO	57.9	48.6	37.8	89.4 (42.2)	89.2 (62.1)
(𝑗=15)	(27.1)	(23.7)	(20)	0.9/9.6/0.1	6.3/2.5/2
Ridge	52.3	44.1	35	91 (58.7)	88 (63)
(𝜆=5)	(22.8)	(19.2)	(17.7)	0.1/8.9/0	4.9/4.8/2.3
EN	51.8	43.9	34.1	92.3 (59.9)	88.6 (65)
(𝜆=0.01; 𝛼=0.4)	(24.3)	(20.5)	(17.2)	0.1/7.6/0	4.9/4.4/2.1