A New Generative Neural Network for Bearing Fault Diagnosis with Imbalanced Data

Table 5

Performance of different deep learning methods in ten-classes diagnosis.

Ten classes		Precision	Recall	F1-score	AUC
GCNN	Case 2	0.893	0.874	0.883	0.996
	Case 3	0.942	0.896	0.919	0.996
	Case 4	0.812	0.740	0.774	0.985
	Case 5	0.903	0.797	0.846	0.985
	Case 6	0.912	0.755	0.826	0.970
	Case 7	0.862	0.743	0.798	0.967
WCNN	Case 2	0.917	0.907	0.912	0.986
	Case 3	0.932	0.921	0.926	0.985
	Case 4	0.833	0.788	0.810	0.951
	Case 5	0.872	0.852	0.862	0.959
	Case 6	0.870	0.855	0.862	0.945
	Case 7	0.833	0.792	0.812	0.958
NCNN	Case 2	0.918	0.863	0.890	0.988
	Case 3	0.901	0.829	0.863	0.986
	Case 4	0.865	0.788	0.824	0.974
	Case 5	0.880	0.805	0.841	0.970
	Case 6	0.800	0.691	0.741	0.922
	Case 7	0.734	0.574	0.644	0.926
FNN	Case 2	0.642	0.522	0.576	0.829
	Case 3	0.640	0.543	0.587	0.837
	Case 4	0.542	0.438	0.485	0.777
	Case 5	0.563	0.453	0.502	0.814
	Case 6	0.563	0.409	0.474	0.783
	Case 7	22	0.485	0.544	0.812