Research Article
Comparison of Machine Learning Classification Methods for Determining the Geographical Origin of Raw Milk Using Vibrational Spectroscopy
Table 3
Statistical parameters obtained by the SVM and PCA-SVM methods.
| Confusion matrix (SVM) | Actual | Sensitivity (%) | Specificity (%) | % CCR | R1 | R2 | R3 | R4 |
| Raw | Predicted | R1 | 22 | 0 | 8 | 15 | 48.89 | 96.3 | 66,97 | R2 | 0 | 24 | 0 | 0 | 100 | 100 | R3 | 2 | 0 | 20 | 2 | 46.51 | 87.1 | R4 | 0 | 0 | 0 | 8 | 100 | 79,52 | Detrend polynomial 1 | R1 | 0 | 0 | 0 | 0 | 0 | 69,23 | 69,68 | R2 | 0 | 24 | 0 | 0 | 100 | 100 | R3 | 24 | 0 | 28 | 23 | 100 | 100 | R4 | 0 | 0 | 0 | 2 | 100 | 69,33 | Detrend polynomial 2 | R1 | 0 | 0 | 0 | 0 | 0 | 68,42 | 67,97 | R2 | 0 | 24 | 0 | 0 | 100 | 100 | R3 | 24 | 0 | 28 | 25 | 100 | 100 | R4 | 0 | 0 | 0 | 0 | 0 | 67,53 | Confusion matrix (PCA-SVM) | Actual | Sensitivity (%) | Specificity (%) | % CCR | R1 | R2 | R3 | R4 | Raw | Predicted | R1 | 24 | 0 | 0 | 0 | 100 | 98,68 | 98,51 | R2 | 0 | 24 | 0 | 0 | 100 | 100 | R3 | 1 | 0 | 28 | 2 | 100 | 100 | R4 | 0 | 0 | 0 | 23 | 100 | 97,44 | Detrend polynomial 1 | R1 | 24 | 0 | 0 | 0 | 100 | 100 | 98,49 | R2 | 0 | 24 | 0 | 0 | 100 | 100 | R3 | 0 | 0 | 28 | 3 | 100 | 100 | R4 | 0 | 0 | 0 | 22 | 100 | 96,2 | Detrend polynomial 2 | R1 | 24 | 0 | 0 | 0 | 100 | 100 | 99 | R2 | 0 | 24 | 0 | 0 | 100 | 100 | R3 | 0 | 0 | 28 | 2 | 100 | 100 | R4 | 0 | 0 | 0 | 23 | 100 | 97,44 |
|
|