Comparison of Machine Learning Classification Methods for Determining the Geographical Origin of Raw Milk Using Vibrational Spectroscopy

Table 4

Performance parameters obtained for the validation of the PLS-DA and PCA-LDA models and SVM and PCA-SVM models constructed using FT-MIR spectral data preprocessed by detrend polynomial degree 1.

Confusion matrix			Actual				Sensitivity (%)	Specificity (%)	% CCR
			R1	R2	R3	R4
PLS-DA	Predicted	R1	12	0	0	0	100	100	100
		R2	0	12	0	0	100	100
		R3	0	0	14	0	100	100
		R4	0	0	0	12	100	100
PCA-LDA		R1	12	0	0	0	100	100	100
		R2	0	12	0	0	100	100
		R3	0	0	14	0	100	100
		R4	0	0	0	12	100	100
SVM		R1	0	0	0	0	0	68,42	66,67
		R2	1	12	1	0	85,71	100
		R3	11	0	13	11	100	92,86
		R4	0	0	0	1	100	69,44
PCA-SVM		R1	11	0	0	0	100	97,3	94,95
		R2	0	12	0	0	100	100
		R3	1	0	12	0	100	94,6
		R4	0	0	2	12	85,71	100