LogCAD: An Efficient and Robust Model for Log-Based Conformal Anomaly Detection

Table 4

Experimental results on HDFS unstable log sequences.

Injection rate (%)	Classifier	Accuracy	Precision	Recall	F1	MCC
5	LR	0.999	0.94	0.997	0.967	0.967
	SVM	0.996	0.849	0.994	0.916	0.917
	NB	0.996	0.849	0.998	0.918	0.919
	LogRobust	0.996	0.990	0.930	0.960	0.965
	CP	0.998	0.946	1	0.972	0.972
	LogCAD	1	0.992	1	0.996	0.996
10	LR	0.998	0.94	0.984	0.961	0.961
	SVM	0.996	0.848	0.989	0.913	0.914
	NB	0.996	0.849	0.993	0.915	0.916
	LogRobust	0.997	0.940	0.990	0.961	0.964
	CP	0.998	0.946	1	0.972	0.972
	LogCAD	0.999	0.996	0.992	0.994	0.994
15	LR	0.998	0.94	0.984	0.961	0.961
	SVM	0.996	0.848	0.989	0.913	0.914
	NB	0.996	0.849	0.999	0.996	0.992
	LogRobust	0.998	0.980	0.910	0.940	0.960
	CP	0.998	0.945	0.992	0.968	0.968
	LogCAD	1	0.992	0.992	0.992	0.992
20	LR	0.998	0.939	0.983	0.960	0.960
	SVM	0.993	0.846	0.987	0.911	0.909
	NB	0.995	0.854	0.979	0.912	0.910
	LogRobust	0.998	0.920	0.970	0.950	0.960
	CP	0.998	0.946	0.996	0.97	0.97
	LogCAD	0.999	0.975	0.996	0.985	0.985