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

Table 2

Random forest, decision tree, AdaBoost, and KNN algorithms for evaluation results of metric combinations. Taking the first row of the random forest algorithm as an example, the vulnerability severity assessment using only one metric, 5, can still achieve an accuracy rate of 65.39%.

Random forest		Decision tree		AdaBoost		KNN
Metric combination	Accuracy	Metric combination	Accuracy	Metric combination	Accuracy	Metric combination	Accuracy
5	0.6539	5	0.6539	5	0.6539	5	0.6539
6	0.6687	6	0.6687	6	0.6687	7	0.6523
0, 6	0.6749	0, 6	0.6749	0, 6	0.6749	0, 7	0.6581
5, 6	0.6765	5, 6	0.6765	5, 6	0.6765	3, 5	0.6578
2, 3, 5	0.7785	2, 3, 5	0.7785	2, 3, 5	0.7785	0, 2, 6	0.7637
2, 3, 6	0.8238	2, 3, 6	0.8238	2, 3, 6	0.7694	2, 3, 6	0.8239
2, 3, 5, 6	0.8894	2, 3, 5, 6	0.8894	2, 3, 5, 6	0.8882	2, 3, 5, 6	0.8804
2, 3, 5, 7	0.8841	2, 3, 5, 7	0.8841	2, 3, 5, 7	0.8841	2, 3, 5, 7	0.8646
0, 2, 3, 5, 6	0.9280	0, 2, 3, 5, 6	0.9281	0, 2, 3, 5, 7	0.9225	0, 2, 3, 5, 6	0.9262
0, 2, 3, 5, 7	0.9225	0, 2, 3, 5, 7	0.9225	1, 2, 3, 5, 7	0.9123	0, 2, 3, 5, 7	0.9219
0, 1, 2, 3, 5, 6	0.9574	0, 1, 2, 3, 5, 6	0.9575	0, 1, 2, 3, 5, 6	0.9527	0, 1, 2, 3, 5, 6	0.9560
0, 2, 3, 5, 6, 7	0.9525	0, 2, 3, 5, 6, 7	0.9523	0, 1, 2, 3, 5, 7	0.9480	0, 2, 3, 5, 6, 7	0.9493
0, 1, 2, 3, 4, 5, 6	0.9699	0, 1, 2, 3, 4, 5, 6	0.9699	0, 1, 2, 3, 5, 6, 7	0.9850	0, 1, 2, 3, 4, 5, 6	0.9685
0, 1, 2, 3, 5, 6, 7	0.9852	0, 1, 2, 3, 5, 6, 7	0.9849	0, 2, 3, 4, 5, 6, 7	0.9640	0, 1, 2, 3, 5, 6, 7	0.9844