Table 3.
Main study: Mean ± standard deviation of classification performance metrics over the 5-fold model evaluation CV with features separately computed over the Interview and Mental Arithmetics phases of the (f-)TSST, respectively.
| Scaler | Feature selection | Classifier | Accuracy [%] | F1-score [%] | Precision [%] |
|---|---|---|---|---|---|
| Min-Max | SFM | RF | 73.4 (7.7) | 71.7 (9.7) | 75.4 (7.6) |
| Min-Max | SkB | DT | 70.7 (8.8) | 69.9 (12.5) | 70.3 (8.8) |
| Min-Max | SkB | NB | 68.0 (6.3) | 64.2 (10.1) | 71.6 (6.6) |
| Standard | SkB | MLP | 68.0 (6.3) | 63.2 (13.8) | 71.2 (2.7) |
| Standard | SkB | kNN | 67.0 (6.3) | 59.1 (9.9) | 76.0 (5.6) |
| Min-Max | RFE | SVM | 66.8 (5.4) | 64.3 (8.7) | 68.3 (3.3) |
| Standard | RFE | Ada | 63.0 (6.3) | 63.5 (3.8) | 64.5 (9.7) |
For each evaluated classifier, the classification pipeline combination with the highest mean accuracy is shown. The classification pipelines scoring the highest metrics are highlighted in bold.