Table 5 Performances comparison between 3 models of each quality indices.

Parameters	Methods	PCs/c	N/g	Calibration set			Prediction set
Parameters	Methods	PCs/c	N/g	Rc	RMSEC	Bias	Rp	RMSEP	Bias	SEP	CV	RPD
TFs	PLS	6	—	0.811	0.068	−0.001	0.795	0.075	−0.013	0.013	0.187	1.206
	SVM	0.04^a	0.32^b	0.881	0.055	0.001	0.886	0.056	−0.004	0.012	0.220	1.526
	RF	5	700	0.970	0.033	−0.001	0.891	0.058	−0.007	0.011	0.190	1.612
TRs	PLS	7	—	0.733	0.368	−0.006	0.752	0.388	−0.041	0.053	0.079	0.936
	SVM	5.38^a	0.066^b	0.853	0.282	−0.008	0.838	0.326	−0.015	0.046	0.085	1.212
	RF	7	100	0.969	0.163	0.000	0.890	0.297	−0.027	0.044	0.081	1.267
TBs	PLS	5	—	0.936	0.347	0.001	0.921	0.398	−0.010	0.101	0.137	2.346
	SVM	5.91^a	0.065^b	0.964	0.268	−0.049	0.940	0.294	−0.049	0.101	0.136	2.511
	RF	2	200	0.986	0.168	0.003	0.944	0.347	−0.068	0.101	0.136	2.636
Sensory score	PLS	5	—	0.929	1.887	−0.006	0.936	1.843	−0.179	0.535	0.059	2.564
	SVM	26.37^a	3.39^b	0.959	1.428	−0.094	0.941	1.670	0.347	0.600	0.065	2.897
	RF	2	700	0.987	0.867	0.032	0.948	1.733	0.482	0.571	0.062	2.931

^aRepresents penalty parameters (c) of SVM model; ^bis the kernel function parameters c of SVM model.
SD, standard deviation; PCs, used latent variables; RMSEC, root mean square error of calibration; RMSEP: root mean square error of prediction; SEP, standard error of prediction; RPD, residual predictive deviation value of prediction.

Quick links

Search