Table 3

Performance metrics on each machine learning approach with versus without using Synthetic Minority Oversampling Technique

Machine learning approach	F1		Accuracy		Precision		Recall		AUC
Machine learning approach	No SMOTE	SMOTE	No SMOTE	SMOTE	No SMOTE	SMOTE	No SMOTE	SMOTE	No SMOTE	SMOTE
Logistic regression	0.473	0.542	0.806	0.749	0.643	0.473	0.379	0.644	0.794	0.766
Balanced random forest classifier	0.747	0.847	0.863	0.933	0.656	0.891	0.874	0.813	0.936	0.959
Balanced bagging classifier	0.803	0.841	0.901	0.931	0.752	0.887	0.869	0.806	0.942	0.959
Random forest classifier	0.797	0.847	0.919	0.932	0.934	0.884	0.701	0.818	0.957	0.959
Multilayer perceptron classifier	0.399	0.505	0.802	0.712	0.690	0.440	0.301	0.638	0.777	0.759
Support vector classifier	0.475	0.449	0.724	0.653	0.436	0.603	0.531	0.603	0.727	0.707

Values in green font signify improvement in given metric when SMOTE is used. Values in red font signify decrease in performance of given metric when SMOTE is used.
AUC, area under the curve; SMOTE, Synthetic Minority Oversampling Technique.