Nondestructive Discrimination of Lead (Pb) in Preserved Eggs (Pidan) by Near-Infrared Spectroscopy and Chemometrics

Table 2

Discrimination of unleaded (positive) and leaded (negative) pidan objects by PLSDA and LS-SVM.

Methods	Sensitivity	Specificity	MRMCCV	Accuracy	Parameters
Raw data, PLSDA	0.925 (74/80^a)	0.938 (30/32^b)	0.170	0.929 (104/112^c)	7^d
Smoothing, PLSDA	0.913 (73/80)	0.906 (29/32)	0.191	0.911 (102/112)	7
2nd derivative, PLSDA	0.975 (78/80)	1.000 (32/32)	0.027	0.982 (110/112)	5
SNV, PLSDA	0.950 (76/80)	0.938 (30/32)	0.091	0.946 (106/112)	6
Raw data, LS-SVM	0.900 (72/80)	0.906 (29/32)	0.178	0.902 (101/112)	0.55, 12^e
Smoothing, LS-SVM	0.938 (75/80)	0.875 (28/32)	0.103	0.920 (103/112)	0.40, 14
2nd derivative, LS-SVM	0.975 (78/80)	0.969 (31/32)	0.038	0.973 (109/112)	0.80, 12
SNV, LS-SVM	0.950 (76/80)	0.969 (31/32)	0.069	0.955 (107/112)	0.25, 7

aTrue positive/total positive.
^bTrue negative/total negative.
^cThe number of correctly classified/the total number of test objects.
^dThe number of PLS components.
^e𝜎² and 𝛾 for LS-SVM.