Table 1 Identification results of LDA, BP-ANN and SVM models

Models	Subsets	Sample type	Sample number	Discrimination results
Models	Subsets	Sample type	Sample number	Fresh	Secondary fresh	Stale	Discrimination rate (%)
LDA	Training set	Fresh	20	20	0	0	91.67^a
		Secondary fresh	20	1	16	3
		Stale	20	0	1	19
	Prediction set	Fresh	10	10	0	0	93.33^b
		Secondary fresh	10	1	8	1
		Stale	10	0	0	10
BP-ANN	Training set	Fresh	20	20	0	0	95.00^c
		Secondary fresh	20	1	18	1
		Stale	20	0	1	19
	Prediction set	Fresh	10	10	0	0	100
		Secondary fresh	10	0	10	0
		Stale	10	0	0	10
SVM	Training set	Fresh	20	20	0	0	98.33^d
		Secondary fresh	20	1	19	0
		Stale	20	0	0	20
	Prediction set	Fresh	10	10	0	0	93.33^e
		Secondary fresh	10	1	8	1
		Stale	10	0	0	10

^a LDA model in the training set: one secondary fresh sample was misclassified as fresh meat, three secondary fresh samples were misclassified as stale meat, one stale sample was misclassified as secondary meat.
^b LDA model in the prediction set: one secondary fresh sample was misclassified as fresh meat, one secondary fresh sample was misclassified as stale meat.
^c BP-ANN model in the training set: one secondary fresh sample was misclassified as fresh meat, one secondary fresh sample was misclassified as stale meat, one stale sample was misclassified as secondary fresh meat.
^d SVM model in the training set: one secondary fresh sample was misclassified as fresh meat.
^e SVM model in the prediction set: one secondary fresh sample was misclassified as fresh meat, one secondary fresh sample was misclassified as stale meat.
LDA, linear discriminant analysis; BP-ANN, back propagation artificial neural network; SVM, support vector machine.