Table 1 Evaluation schemas.
Scenarios | Binary stress detection | Stress level prediction | Stress level prediction with cold start |
|---|---|---|---|
Labels | Not stressed, stressed | Below median, median, above median | Below median, median, above median |
Evaluations | Five-fold cross-validation | Five-fold cross-validation | Leave-one-subject-out validation |
Sections | “CALM-Net and branched CALM-Net attain AUC scores of more than 0.8 on student stress detection” section. On binary stress detection, CALM-Net attains precisions higher than \(84\%\) when recovering \(90\%\) positive cases | “Branched CALM-Net improves upon the SOTA on 3-class stress level prediction” section. Branched CALM-Net improves the state-of–the-art by introducing personalization and dynamic clustering on stress level prediction with 3 classes | “With 1 week of data, branched CALM-Net achieves 17.24% boost in performance over SOTA” section. Branched CALM-Net is the top performer with F1 score of 0.67 when training on 1 week data from new subject |
