Table 1.
Model optimization using the classifier algorithms. Log loss values are means of 10-fold cross-validation.
| Algorithm | Tuning Parameter | Mean Log Loss Values |
|---|---|---|
| KKNN | Kmax = 5 | 0.8339319 |
| Kmax = 7 | 0.9017721 | |
| Kmax = 9 | 0.9808674 | |
| PDA | Lambda = 1 | 0.5689435 |
| Lambda = 0.0001 | 0.5687306 | |
| Lambda = 0.1 | 0.4611719 | |
| HDDA | Thershold = 0.05 | 0.4360396 |
| Thershold = 0.175 | 1.3732500 | |
| Thershold = 0.300 | 1.0080708 | |
| SDA | Lambda = 0.0 | 0.6320813 |
| Lambda = 0.5 | 0.3968958 | |
| Lambda = 1.0 | 0.4908678 | |
| PAM | Threshold = 0.7608929 | 0.4986565 |
| Threshold = 11.0329476 | 1.9483062 | |
| Threshold = 21.3050022 | 1.9483062 | |
| PLS | Ncomp = 1 | 1.826913 |
| Ncomp = 2 | 1.733439 | |
| Ncomp = 3 | 1.643669 | |
| C5.0 tree | 0.7482527 | |
| ET | 0.3590714 |
KKNN: weighted k-nearest neighbors; PDA: penalized discriminant analysis; HDDA: high-dimensional discriminant analysis; SDA: shrinkage discriminant analysis; PAM: nearest shrunken centroids; PLS: partial least squares; ET: extremely randomized trees.