Table 4. Performance of 3 models with and without SMOTE.

Model	Evaluation	KDDCup99		CICIDS 2017		UNSW_NB15
		Without SMOTE	With SMOTE	Without SMOTE	With SMOTE	Without SMOTE	With SMOTE
LightGBM	accuracy	0.878	0.977	0.653	0.996	0.845	0.893
	Precision	0.920	0.975	0.721	0.996	0.857	0.899
	Recall	0.878	0.977	0.653	0.996	0.845	0.893
	F1-Score	0.887	0.976	0.657	0.996	0.850	0.895
XGBoost	accuracy	0.878	0.999	0.998	0.995	0.899	0.893
	Precision	0.920	0.999	0.998	0.995	0.901	0.901
	Recall	0.878	0.999	0.998	0.995	0.899	0.893
	F1-Score	0.887	0.999	0.998	0.994	0.897	0.896
CatBoost	accuracy	1.000	0.999	0.997	0.996	0.892	0.887
	Precision	1.000	0.999	0.997	0.996	0.893	0.889
	Recall	1.000	0.999	0.997	0.996	0.892	0.887
	F1-Score	1.000	0.999	0.997	0.996	0.888	0.887