VulDistilBERT: A CPS Vulnerability Severity Prediction Method Based on Distillation Model

Table 12

The results of comparing the performance of different classifiers on the dataset with the optimal subset integrated and data augmented.

Models	Classifiers	Best epoch	Loss	Accuracy	Precision	Recall	F1 scores
Pretrained DistilBERT	Linear	34	0.814	0.75	0.77	0.76	0.76
Fine-tuned DistilBERT	Linear	2	0.095	0.97	0.97	0.97	0.97
Fine-tuned DistilBERT	CNN	2	0.093	0.97	0.97	0.97	0.97
Fine-tuned DistilBERT	LSTM	2	0.092	0.97	0.97	0.97	0.97