Deep Forest-Based Fault Diagnosis Method for Chemical Process

Table 5

FDR obtained by RF, XGBoost, AdaBoost, gcForest, and WCForest.

FDR (%)	RF	XGBoost	AdaBoost	gcForest	WCForest	Percentage increase of WCForest compared to RF
Fault 1	100	100	1.83	99.42	99.17	−0.83
Fault 2	100	100	97.83	98.17	100	0
Fault 3	15.17	16.5	0	41.59	42.33	27.16
Fault 4	99.17	98.83	0	98.26	98.83	−0.34
Fault 5	28.83	99.67	50.67	83.74	90.97	62.14
Fault 6	100	100	0	100	100	0
Fault 7	100	100	0	100	100	0
Fault 8	62.50	69.17	52.67	93.36	96	33.50
Fault 9	5.83	5	0	53.18	51.50	45.67
Fault 10	30.50	40.67	9.17	78.21	75.83	45.33
Fault 11	61.33	67.67	49.33	83.45	82.00	20.67
Fault 12	80.00	77.83	29	87.29	93.50	13.50
Fault 13	45.67	50.33	26.67	89.21	94.83	49.16
Fault 14	94.83	96.67	61.33	97.03	96.83	2.00
Fault 15	0.83	3.67	0	6.48	7.17	6.34
Fault 16	35.00	49.67	0	72.37	78.00	43.00
Fault 17	87.17	87.83	42.17	94.00	94.50	7.33
Fault 18	99.33	99.83	0.5	90.84	99.67	0.34
Fault 19	50.50	58	2.33	85.32	89.50	39.00
Fault 20	61.00	67.83	37.5	76.05	76.00	15.00
Fault 21	0.67	46.33	0	98.87	100	99.33
Average	59.92	68.36	21.95	82.23	84.13	24.21