Table 2.
Method | Precision | Recall | F-score |
---|---|---|---|
Direct combination | |||
LR | 0.836 | 0.822 | 0.823 |
RF | 0.869 | 0.865 | 0.865 |
SVM | 0.833 | 0.830 | 0.827 |
Auto-encoder + LR | 0.664 | 0.650 | 0.651 |
Auto-encoder + RF | 0.787 | 0.783 | 0.777 |
Auto-encoder + SVM | 0.767 | 0.767 | 0.757 |
Multimodal deep auto-encoder | |||
Multimodal auto-encoder + LR | 0.841 | 0.832 | 0.831 |
Multimodal auto-encoder + RF | 0.883 | 0.881 | 0.879 |
Multimodal auto-encoder + SVM | 0.860 | 0.857 | 0.854 |
Stacked ensemble | |||
MEL-MP | 0.895 | 0.893 | 0.892 |
Each feature fusion method is shown in bold, and the experimental results of MEL-MP are also shown in bold.