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

EPJ Data Science

Table 15 Classification results for the bank marketing dataset

Evaluation metric	Naive Bayes	SGD	KNN	MLP	Random forest	SVM
Without augmentation
F1 score macro	0.3734	0.405	0.4312	0.3517	0.3791	0.4338
F1 score micro	0.557	0.5405	0.5753	0.5464	0.5518	0.5524
Precision minority class	0	0.4347	0.8308	0	0.5362	0.5464
Recall minority class	0.0222	0.1129	0.0862	0.0011	0.0308	0.1106
Baseline: min-max approach
F1 score macro	0.3738	0.4475	0.5326	0.311	0.477	0.3083
F1 score micro	0.3824	0.5413	0.5714	0.419	0.5836	0.454
Precision minority class	0.3726	0.5074	0.5274	0.4331	0.5569	0.454
Recall minority class	0.5384	0.1496	0.6065	0.8976	0.257	1
Baseline: mean-std approach
F1 score macro	0.3116	0.3504	0.5324	0.5002	0.4594	0.3083
F1 score micro	0.3226	0.4935	0.5713	0.5214	0.5259	0.454
Precision minority class	0.2815	0.1737	0.5273	0.4882	0.5107	0.454
Recall minority class	0.3576	0.0282	0.6062	0.7362	0.5497	1
SMOTE approach
F1 score macro	0.8055	0.4632	0.9105	0.7237	0.9957	0.5509
F1 score micro	0.82	0.567	0.9149	0.7534	0.9957	0.5628
Precision minority class	0.8749	0.4522	0.8525	0.8819	0.9997	0.5187
Recall minority class	0.7024	0.2902	0.9679	0.5326	0.9917	0.5115
SMOTE-clustering approach
F1 score macro	0.8058	0.6152	0.8916	0.7269	0.9972	0.548
F1 score micro	0.82	0.6742	0.8968	0.7546	0.9972	0.5574
Precision minority class	0.8723	0.7603	0.828	0.8784	0.9992	0.5107
Recall minority class	0.707	0.4435	0.9575	0.5346	0.9952	0.5444