Table 1.
Comparative of machine learning algorithms. We used the metrics accuracy and Kappa, for comparing different algorithms in group 1 or constitutive scent profile (pFreqCutoff = 1.0) and group 2 (pFreqCutoff = 0.7). k-NN: k-Nearest Neighbors, NBC: Naïve Bayes Classifier, SVM: Support Vector Machine, RF: Random Forest, SD: standard deviation.
| Metric | Algorithm | Group 1 | Group 2 | ||
|---|---|---|---|---|---|
| Mean | SD | Mean | SD | ||
| Accuracy | k-NN | 0.77 | 0.17 | 0.9 | 0.14 |
| NBC | 0.665 | 0.30 | 0.89 | 0.14 | |
| SVM Linear | 0.79 | 0.14 | 0.92 | 0.14 | |
| SVM Radial | 0.755 | 0.19 | 0.88 | 0.11 | |
| SVM Polynomial | 0.86 | 0.17 | 0.92 | 0.14 | |
| RF | 0.84 | 0.18 | 0.98 | 0.06 | |
| Kappa | k-NN | 0.47 | 0.38 | 0.78 | 0.34 |
| NBC | 0.52 | 0.37 | 0.75 | 0.32 | |
| SVM Linear | 0.53 | 0.40 | 0.82 | 0.34 | |
| SVM Radial | 0.52 | 0.39 | 0.78 | 0.18 | |
| SVM Polynomial | 0.65 | 0.42 | 0.81 | 0.34 | |
| RF | 0.61 | 0.46 | 0.95 | 0.14 |