Leveraging augmentation techniques for tasks with unbalancedness within the financial domain: a two-level ensemble approach

EPJ Data Science

Table 11 Classification results for the bankruptcy in Poland dataset

Evaluation metric	Naive Bayes	SGD	KNN	MLP	Random forest	SVM
Without augmentation
F1 score macro	0.3584	0.4019	0.3583	0.5317	0.3639	0.40
F1 score micro	0.5006	0.5175	0.5692	0.6249	0.5691	0.5161
Precision minority class	0.4997	0	0	0.6824	0	0.6118
Recall minority class	0.9732	0.0868	0	0.2335	0.0062	0.0773
Baseline: min-max approach
F1 score macro	0.3689	0.364	0.5429	0.5243	0.7141	0.3583
F1 score micro	0.4509	0.511	0.6125	0.6023	0.7359	0.5692
Precision minority class	0.4351	0.3145	0.9937	0	0.9966	0
Recall minority class	0.9254	0.0494	0.2247	0.212	0.4725	0
Baseline: mean-std approach
F1 score macro	0.3691	0.3708	0.3583	0.4868	0.7241	0.3583
F1 score micro	0.4513	0.5468	0.5692	0.6052	0.7449	0.5692
Precision minority class	0.4349	0.204	0	0.6491	0.9976	0
Recall minority class	0.9236	0.0284	0	0.1744	0.4815	0
SMOTE approach
F1 score macro	0.3157	0.5932	0.9013	0.8829	0.9381	0.6826
F1 score micro	0.4335	0.6028	0.9137	0.8948	0.9474	0.7029
Precision minority class	0.4305	0.5551	0.8292	0.8541	0.9074	0.6166
Recall minority class	0.9738	0.6581	0.9694	0.8782	0.844	0.6975
SMOTE-clustering approach
F1 score macro	0.4168	0.4494	0.9176	0.6447	0.9082	0.6772
F1 score micro	0.5801	0.4603	0.9293	0.6834	0.9169	0.6949
Precision minority class	0.4537	0.4345	0.8394	0.6814	0.881	0.612
Recall minority class	0.5098	0.6437	0.9836	0.4935	0.9144	0.7091