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

Table 6

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

Four		Precision	Recall	F1-score	AUC
GCNN	Case 2	0.958	0.919	0.938	0.996
	Case 3	0.956	0.956	0.956	0.995
	Case 4	0.984	0.984	0.984	1.000
	Case 5	0.870	0.839	0.855	0.988
	Case 6	0.833	0.800	0.816	0.971
	Case 7	0.872	0.872	0.872	0.975
WCNN	Case 2	0.931	0.905	0.918	0.994
	Case 3	0.776	0.765	0.911	0.928
	Case 4	0.852	0.839	0.846	0.943
	Case 5	0.821	0.821	0.821	0.886
	Case 6	0.720	0.720	0.720	0.869
	Case 7	0.723	0.723	0.723	0.922
NCNN	Case 2	0.857	0.811	0.833	0.972
	Case 3	0.828	0.779	0.803	0.939
	Case 4	0.788	0.661	0.719	0.913
	Case 5	0.846	0.786	0.815	0.932
	Case 6	0.767	0.660	0.710	0.858
	Case 7	0.756	0.660	0.705	0.907
FNN	Case 2	0.662	0.608	0.634	0.744
	Case 3	0.567	0.559	0.563	0.702
	Case 4	0.672	0.661	0.667	0.771
	Case 5	0.589	0.589	0.589	0.710
	Case 6	0.587	0.540	0.563	0.725
	Case 7	0.674	0.660	0.667	0.753