A Comparative Study of Attention-Based Transformer Networks and Traditional Machine Learning Methods for Toxic Comments Classification
Table 1.
Metrics Performance of Attention-Based Transformer andTraditional Machine Learning Methods
|
| Model |
Accuracy |
Precision |
Recall |
F1-Score |
|
|
| Logistic Regression |
85.25% |
82.19% |
88.34% |
85.15% |
|
| Naive Bayes |
78.42% |
75.61% |
82.39% |
78.05% |
|
| Support Vector Machines (SVM) |
87.62% |
84.28% |
89.46% |
86.14% |
|
| Decision Trees |
80.92% |
78.12% |
85.92% |
80.89% |
|
| Random Forests |
86.79% |
83.47% |
88.61% |
85.23% |
|
| Gradient Boosting |
88.45% |
85.92% |
90.11% |
87.38% |
|
| Attention-Based Transformer |
92.14% |
89.28% |
94.13% |
91.12% |
|
| Transformer |
90.68% |
87.83% |
92.04% |
89.19% |
|
|
|
|
