Table 3.

Comparison of algorithm performance with different unbalanced dataset processing strategies.

	TPR	TNR	Pre	ACC	F1	MCC	AUROC	AP
Dset_448
None	0.197	0.979	0.601	0.873	0.297	0.293	0.807	0.433
Cost-Sensitive	0.491	0.898	0.431	0.843	0.459	0.369	0.809	0.430
Ensemble	0.452	0.922	0.477	0.858	0.464	0.383	0.820	0.460
UnderSampler	0.652	0.798	0.336	0.778	0.444	0.350	0.809	0.435
OverSampler	0.549	0.870	0.398	0.826	0.462	0.381	0.807	0.439
Dset_72
None	0.209	0.963	0.398	0.883	0.274	0.230	0.779	0.305
Cost-Sensitive	0.386	0.905	0.325	0.850	0.353	0.270	0.775	0.304
Ensemble	0.318	0.938	0.377	0.872	0.345	0.276	0.788	0.330
UnderSampler	0.587	0.783	0.243	0.762	0.343	0.261	0.769	0.291
OverSampler	0.464	0.870	0.298	0.827	0.363	0.271	0.779	0.315
Dset_164
None	0.134	0.978	0.570	0.825	0.217	0.213	0.741	0.404
Cost-Sensitive	0.418	0.868	0.412	0.787	0.415	0.284	0.744	0.404
Ensemble	0.323	0.916	0.460	0.809	0.380	0.277	0.755	0.413
UnderSampler	0.565	0.761	0.343	0.725	0.427	0.274	0.738	0.378
OverSampler	0.452	0.850	0.401	0.778	0.425	0.281	0.742	0.404