Performance Comparison Table
Binary Classification 2 labels: Bully (sexual, political, threat, troll), and Neutral
| Models | Accuracy | F1 |
|---|---|---|
| Decision Tree | 82% | 54% |
| Naive Bayes | 65% | 65% |
| KNN | 82% | 57% |
| Random Forest | 80% | 44% |
| RNN | 79.70% | 88.47% |
| LSTM | 80.12% | 88.99% |
| GRU | 80.12% | 88.99% |
| Bangla Bert | 89.49% | 93.48% |
Multi class classification (5 labels: sexual, political, threat, troll, neutral)
| Models | Accuracy | F1 |
|---|---|---|
| Decision Tree | 22% | 13% |
| Naive Bayes | 34% | 33% |
| KNN | 29% | 20% |
| Random Forest | 32% | 31% |
| RNN | 23.38% | |
| LSTM | 19.88% | |
| GRU | 19.22% | |
| Bangla Bert | 73.28% | 72.78% |