Qualitative Discrimination of Intact Tobacco Leaves Based on Near-Infrared Technology

Table 4

Results of the qualitative discrimination model of intact tobacco leaves (spectra wavenumbers from 5006–8988 cm⁻¹; preprocessing method: 1st derivative + SNV).

Modeling algorithm	Modeling spectra	Position-GV discrimination		Grade discrimination
		Training set (%)	Validation set (%)	Training set (%)	Validation set (%)
DPLS	Independent spectra	100.00	96.22	90.51	86.25
	Average spectra	98.12	92.17	86.84	81.75
Fisher	Independent spectra	97.67	94.27	99.64	97.49
	Average spectra	100.00	91.43	100.00	91.43