Table 5 Performance metrics of different models
Model | Accuracy | PPV | Sensitivity | F1 score |
|---|---|---|---|---|
XLNet30 | 58.76% | 54.38% | 58.76% | 56.07% |
ALBERT31 | 69.07% | 47.71% | 69.07% | 56.44% |
DistilBERT32 | 58.76% | 51.85% | 58.76% | 54.44% |
RoBERTa33 | 57.73% | 58.12% | 57.73% | 57.92% |
BERT34 | 63.92% | 60.86% | 63.92% | 61.87% |
Mistral - 7b | 70.29% | 70.29% | 100.00% | 82.55% |
Qwen2 - 7b | 70.29% | 70.29% | 100.00% | 82.55% |
Gemini 2.0 Flash-based | 63.47% | 72.34% | 77.97% | 75.05% |
GPT 4o - based | 67.05% | 70.76% | 90.51% | 79.43% |
GPT 3.5 - based | 69.14% | 71.52% | 93.22% | 80.94% |
