Table 8.
Comparison of proposed ML classifiers accuracy with baseline using word embeddings.
| Model dataset | COVIDSENTI | Covidsenti_A | Covidsenti_B | CovidSenti_C | ||
|---|---|---|---|---|---|---|
| Existing models/Accuracy | ||||||
| Word2Vec | RF | 76.09% | 76.04% | 73.02% | 75.03% | |
| DT | 76.06% | 76.02% | 74.05% | 74.01% | ||
| Glove | RF | 72.06% | 71.06% | 70.02% | 72.02% | |
| DT | 70.01% | 69.03% | 68.09% | 69.04% | ||
| FastText | SVM | 81.05% | 80.01% | 79.02% | 78.03% | |
| NB | 73.05% | 73.02% | 74.05% | 72.01% | ||
| RF | 84.05% | 82.03% | 84.01% | 80.02% | ||
| Proposed model/Accuracy | ||||||
| Fine tuned FastText | XGB | 80.00% | 78.33% | 77.61% | 76.30% | |
| Fine tuned Word2Vec | XGB | 79.71% | 78.33% | 77.85% | 76.05% | |
| Fine tuned Glove | XGB | 79.19% | 78.39% | 78.23% | 76.05% | |
| Loss | 4.05% | 3.64% | 5.78% | 3.97% | ||
The bold ones represent better results.