TABLE 1.

Authenticity identification results and model parameters [Se = TP/(TP + FN); Sp = TN/(TN + FP); Ac = TP/(TP + FP); TP: true positive; FP: false positive; TN: true negative; FN: false negative].

Model	Data matrix	Calibration set					Validation set
		Misclassified samples	Not-assigned samples	Se	Sp	Ac	Misclassified or not-assigned samples	Ac
PLS-DA	EN	3	0	0.9400	0.9800	0.9700	1	0.9706
	EE	2	0	1.0000	0.9700	0.9800	1	0.9706
	ET	1	0	0.9700	1.0000	0.9900	1	0.9706
	HPLC	2	0	0.9700	0.9800	0.9800	0	1.0000
	Data fusion by PCs	0	0	1.0000	1.0000	1.0000	1	0.9706
	Data fusion by LVs	0	0	1.0000	1.0000	1.0000	0	1.0000
SVM	EN	8	0	0.9100	0.9200	0.9200	4	0.8824
	EE	1	0	1.0000	0.9800	0.9900	1	0.9706
	ET	10	0	1.0000	0.9800	0.9000	2	0.9412
	HPLC	6	0	0.8500	0.9800	0.9400	0	1.0000
	Data fusion by PCs	0	0	1.0000	1.0000	1.0000	1	0.9706
	Data fusion by LVs	0	0	1.0000	1.0000	1.0000	0	1.0000
BP-NN	EN	2	0	0.9400	1.0000	0.9800	1	0.9706
	EE	1	0	1.0000	0.9800	0.9900	1	0.9706
	ET	1	1	1.0000	0.9800	0.9800	1	0.9706
	HPLC	5	0	0.9100	0.9700	0.9500	0	1.0000
	Data fusion by PCs	1	0	1.0000	0.9800	0.9900	1	0.9706
	Data fusion by LVs	1	0	1.0000	0.9800	0.9900	1	0.9706

The bold values indicate the optimal model.