Table 9. Experiment results of FastText embedding model including various N-grams with deep learning algorithms for independent dataset.
The best accuracy percentages are represented in bold letters.
| MODELS | ACC | SEN | SPE | MCC | AUC |
|---|---|---|---|---|---|
| FT(2)+CNN | 92.18 | 92.45 | 91.76 | 84.02 | 96.83 |
| FT(3)+CNN | 93.75 | 93.62 | 93.18 | 87.56 | 97.54 |
| FT(4)+CNN | 57.81 | 59.82 | 53.64 | 15.21 | 62.43 |
| FT(2)+LSTM | 87.50 | 88.24 | 86.36 | 73.86 | 91.57 |
| FT(3)+LSTM | 89.06 | 87.32 | 90.12 | 76.03 | 92.74 |
| FT(4)+LSTM | 62.33 | 56.34 | 59.28 | 16.07 | 58.89 |
| FT(2)+BiLSTM | 96.15 | 92.80 | 95.65 | 88.85 | 93.20 |
| FT(3)+BiLSTM | 93.85 | 92.87 | 94.63 | 87.41 | 96.85 |
| FT(4)+BiLSTM | 67.20 | 55.79 | 59.83 | 15.74 | 60.29 |