Table 2 Performance comparison of CNN classifiers trained with different feature encodings on the training dataset.
Feature | AUC | ACC (%) | Sen (%) | Spe (%) | MCC | Pre (%) | F1 |
|---|---|---|---|---|---|---|---|
GloVe | 0.810 | 75.00 | 71.15 | 77.63 | 0.485 | 68.52 | 0.698 |
fastText | 0.785 | 73.44 | 67.24 | 78.57 | 0.462 | 72.22 | 0.696 |
Word2Vec | 0.793 | 71.09 | 73.13 | 68.85 | 0.420 | 72.06 | 0.726 |
fastText + Word2Vec | 0.826 | 75.78 | 70.18 | 80.28 | 0.508 | 74.07 | 0.721 |
GloVe + Word2Vec | 0.820 | 76.64 | 71.96 | 85.00 | 0.523 | 77.50 | 0.713 |
GloVe + fastText | 0.839 | 78.12 | 72.96 | 89.19 | 0.549 | 80.95 | 0.738 |
GloVe + fastText +Word2Vec | 0.864 | 79.07 | 75.10 | 91.49 | 0.585 | 86.21 | 0.740 |
