Table 3.

The results of PLS-DA and SVM models using optimal wavelengths of three sample sets.

Model	Sample	Methods	Parameters	Calibration Accuracy (%)	Prediction Accuracy (%)
PLS-DA	Set 1	PCA loadings	5	76.47	81.82
		Second derivative	7	77.94	68.18
		CARS	14	100.00	72.73
		SPA	14	100.00	82.61
	Set 2	PCA loadings	5	97.78	96.67
		Second derivative	5	93.33	96.67
		CARS	12	100.00	100.00
		SPA	5	93.33	96.67
	Set 3	PCA loadings	5	97.78	100.00
		Second derivative	5	100.00	100.00
		CARS	9	100.00	96.67
		SPA	12	100.00	100.00
SVM	Set 1	PCA loadings	(84.4485, 27.8576)	72.06	77.27
		Second derivative	(0.5743, 256)	69.12	90.91
		CARS	(256,84.4485)	91.18	81.82
		SPA	(147.0334, 256)	94.12	90.91
	Set 2	PCA loadings	(147.0334, 256)	98.89	96.67
		Second derivative	(147.0334, 84.4485)	97.78	100.00
		CARS	(256, 27.8576)	90.00	93.33
		SPA	(256, 147.0334)	100.00	96.67
	Set 3	PCA loadings	(27.8576, 84.4485)	97.78	100.00
		Second derivative	(256, 16)	100.00	100.00
		CARS	(5.2780, 27.8576)	91.11	83.22
		SPA	(5.2780, 84.4485)	98.89	100.00