Table 2.

The classification efficiency values and total accuracy of independent decision making with Partial least squares discriminant analysis (PLS-DA) and random forest (RF) models. RFE: Recursive feature elimination.

	Model	Calibration Set						Validation Set
		Class1	Class2	Class3	Class4	Class5	Accuracy	Class1	Class2	Class3	Class4	Class5	Accuracy
FT-MIR	PLS-DA	0.961	1.000	0.995	0.981	0.990	97.66%	1.000	0.991	0.913	1.000	0.991	97.06%
	RF	0.772	0.888	0.801	0.829	0.790	71.88%	0.886	0.964	0.973	1.000	0.946	92.65%
NIR	PLS-DA	1.000	1.000	1.000	1.000	1.000	100%	0.870	0.964	0.940	0.936	0.964	89.71%
	RF	0.803	0.854	0.775	0.837	0.834	72.66%	0.813	0.917	0.491	0.923	0.955	76.47%
FT-MIR (RFE)	PLS-DA	0.911	0.990	0.881	0.961	0.975	91.41%	0.794	1.000	0.694	0.955	0.923	82.35%
	RF	0.947	0.951	0.853	0.876	0.942	86.72%	0.845	0.917	0.964	0.964	0.936	88.24%
FT-MIR (Bo)	PLS-DA	0.951	0.961	0.995	0.961	0.985	95.31%	0.886	0.991	0.905	0.964	0.962	91.18%
	RF	0.890	0.942	0.829	0.881	0.922	83.59%	0.886	0.964	0.973	1.000	0.946	92.65%
FT-MIR (PCs)	PLS-DA	0.906	0.911	0.868	0.927	0.863	83.59%	0.926	0.991	0.843	0.926	0.972	89.71%
	RF	0.780	0.922	0.730	0.764	0.772	68.75%	0.964	1.000	0.991	0.917	0.991	95.59%
NIR (RFE)	PLS-DA	0.807	0.922	0.926	0.902	0.966	85.16%	0.779	0.845	0.675	0.891	0.953	75%
	RF	0.733	0.888	0.791	0.888	0.942	77.34%	0.813	0.964	0.567	0.889	0.955	77.94%
NIR (Bo)	PLS-DA	0.807	0.922	0.858	0.906	0.947	82.81%	0.779	0.837	0.551	0.962	0.962	75%
	RF	0.729	0.878	0.764	0.893	0.922	75.78%	0.772	0.878	0.486	0.870	0.955	72.06%
NIR (PCs)	PLS-DA	0.860	0.937	0.974	0.915	0.990	89.84%	0.927	0.926	0.991	0.878	0.955	90%
	RF	0.745	0.922	0.881	0.881	0.951	81.25%	0.955	0.964	1.000	1.000	0.991	97.06%

Bo: Boruta, PCs: Principal components.