Quality Degradation of Chinese White Lotus Seeds Caused by Dampening during Processing and Storage: Rapid and Nondestructive Discrimination Using Near-Infrared Spectroscopy

Table 1

The model parameters and prediction results of full-spectrum PLSDA and MWPLSDA models.

Preprocessing	Models	Spectral range (cm^−1)	LVs^a	Sensitivity^b	Specificity^c	Accuracy^d
Raw data	PLSDA	4000–9000	4	0.727 (40/55)	0.775 (31/40)	0.747 (71/95)
Smoothing	PLSDA	4000–9000	4	0.727 (40/55)	0.800 (32/40)	0.758 (72/95)
D2	PLSDA	4000–9000	3	0.782 (43/55)	0.875 (35/40)	0.821 (78/95)
SNV	PLSDA	4000–9000	3	0.800 (44/55)	0.775 (31/40)	0.789 (75/95)
Raw data	MWPLSDA	4904–4991, 4746–4811	3	0.855 (47/55)	0.850 (34/40)	0.853 (81/95)
Smoothing	MWPLSDA	4904–4991, 4746–4811	3	0.855 (47/55)	0.850 (34/40)	0.853 (81/95)
D2	MWPLSDA	7054–7129, 5444–5494, 5234–5348	2	0.873 (48/55)	0.95 (38/40)	0.905 (86/95)
SNV	MWPLSDA	4236–4300, 4908–4979, 7191–7243, 7608–7708, 7943–7992, 8911–8956	3	0.927 (51/55)	0.950 (38/40)	0.937 (89/95)

Number of PLSDA latent variables.
^bThe numbers in the brackets indicate TP/(TP + FN).
^cThe numbers in the brackets indicate TN/(TN + FP).
^dThe numbers in the brackets indicate (TN + TP)/(TN + TP + FP + FN).