Table 2.
The optimized models for the 3-class classification of FT-IR spectral data.
| LDA model for FT-IR | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| calibration: error rate 1%, accuracy 99% | 10-fold CV: error rate 3%, accuracy 97% | Prediction: error rate 4%, accuracy 94% | |||||||
| class | specificity | sensitivity | precision | specificity | sensitivity | precision | specificity | sensitivity | precision |
| beef meat | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% | 100% |
| with beef offal | 100% | 97% | 100% | 98% | 94% | 98% | 92% | 96% | 91% |
| with pork offal | 98% | 100% | 97% | 95% | 98% | 95% | 97% | 90% | 96% |
| # of spectra in calibration dataset for beef meat, with beef offal and with pork offal are 40, 195 and 195, respectively. | |||||||||
| # of spectra in prediction dataset for beef meat, with beef offal and with pork offal are 15, 105 and 105, respectively. | |||||||||