Table 5. The second ablation study, trained on the imbalanced training dataset.
Model | Acc | F1 | Sens | Spec | MCC | AUC |
---|---|---|---|---|---|---|
Validated on the imbalanced validation dataset | ||||||
SSMFN CNN | 0.8939 | 0.8502 | 0.9389 | 0.8834 | 0.7230 | 0.8179 |
SSMFN LSTM | 0.9167 | 0.8891 | 0.9100 | 0.9186 | 0.7836 | 0.8704 |
SSMFN Merged | 0.9078 | 0.8774 | 0.8895 | 0.9133 | 0.7598 | 0.8596 |
Validated on the balanced validation dataset | ||||||
SSMFN CNN | 0.7529 | 0.7372 | 0.9948 | 0.6698 | 0.5798 | 0.8179 |
SSMFN LSTM | 0.8638 | 0.8624 | 0.9567 | 0.8024 | 0.7560 | 0.8704 |
SSMFN Merged | 0.8656 | 0.8640 | 0.9672 | 0.8003 | 0.7491 | 0.8596 |
Tested on the test dataset | ||||||
SSMFN CNN | 0.7404 | 0.7228 | 0.9845 | 0.6598 | 0.5566 | 0.7404 |
SSMFN LSTM | 0.8442 | 0.8418 | 0.9590 | 0.7754 | 0.7110 | 0.8442 |
SSMFN Merged | 0.8462 | 0.8435 | 0.9688 | 0.7744 | 0.7173 | 0.8462 |
Note.
The highest value of each parameter from each measurement experiment is shown in bold.