Table 15 Multimodal results based on combined text-audio feature selectors
From: Multimodal machine learning for language and speech markers identification in mental health
Model | Features | Accuracy | AUC-ROC | F1 - 0s | F1 - 1s |
|---|---|---|---|---|---|
SVM | 20t, 15a | 86.17% | 92.80% | 0.91 | 0.79 |
SVM | 20t, 10a | 86.71% | 92.74% | 0.92 | 0.80 |
SVM | 15t, 15a | 82.97% | 90.61% | 0.89 | 0.72 |
RF | 20t, 15a | 79.80% | 84.45% | 0.87 | 0.60 |
RF | 20t, 10a | 80.87% | 86.39% | 0.87 | 0.63 |
RF | 15t, 15a | 79.82% | 84.38% | 0.87 | 0.60 |
LogReg | 20t, 15a | 85.14% | 91.05% | 0.89 | 0.77 |
LogReg | 20t, 10a | 86.73% | 92.36% | 0.90 | 0.80 |
LogReg | 15t, 15a | 84.57% | 91.01% | 0.89 | 0.74 |
Dense Layers | 20t, 15a | 84.59% | 89.55% | 0.90 | 0.76 |
Dense Layers | 20t, 10a | 84.04% | 90.22% | 0.89 | 0.74 |
Dense Layers | 15t, 15a | 84.07% | 89.91% | 0.88 | 0.70 |