Comparing the Potential of Near- and Mid-Infrared Spectroscopy in Determining the Freshness of Strawberry Powder from Freshly Available and Stored Strawberry

Table 3

Prediction result of strawberry storage day based on the combination of NIRⁿ and MIR^o spectra with full variables and optimal variables.

Model	Variable selection	Variable number	Calibration			Prediction			RPD^l	AB_RMSE^m
					RMSEC^h			RMSEP^k
PLSR^a	Full-variable	3836	0.990	0.980	1.404	0.967	0.934	2.473	3.942	1.069
LS_SVM^b	Full-variable	3836	1.000	1.000	0.002	0.993	0.986	1.152	8.404	1.150
PLSR	SPA^c	6	0.981	0.962	1.930	0.978	0.957	2.007	4.828	0.077
LS_SVM	SPA	6	0.999	0.997	0.505	0.994	0.989	1.027	9.397	0.522
PLSR	UVE^d	731	0.996	0.991	0.920	0.983	0.965	1.818	5.334	0.899
LS_SVM	UVE	731	0.999	0.998	0.469	0.994	0.988	1.051	9.223	0.582
PLSR	UVE_SPA	12	0.994	0.987	1.109	0.976	0.950	2.168	4.525	1.058
LS_SVM	UVE_SPA	12	0.999	0.998	0.462	0.994	0.988	1.051	9.225	0.589
PLSR	CARS^e	44	0.996	0.992	0.860	0.952	0.905	2.982	3.253	2.122
LS_SVM	CARS	44	0.998	0.996	0.603	0.981	0.961	1.906	5.089	1.302

^aPartial least squares regression, ^bleast squares support vector machines, ^csuccessive projections algorithm, ^duninformative variable elimination, ^ecompetitive adaptive reweighted sampling, ^fcorrelation coefficients of calibration, ^gdetermination coefficients of calibration, ^hroot-mean-square error of calibration, ⁱcorrelation coefficients of prediction, ^jdetermination coefficients of prediction, ^kroot-mean-square error of prediction, ^lresidual prediction deviation, ^mabsolute difference between RMSEC and RMSEP, ⁿnear-infrared, and ^omid-infrared.