Table 2.

Performance Results of the Selected Models Developed with Chiang et al. Dataset

Fault	Optimal feature subset size	AUC	Accuracy	Detection rate	False negative rate	False alarm rate	Latency (min)
1*	2	100.0	99.9	99.9	0.1	0.0	6
2*	5	99.5	98.1	97.8	2.3	0.0	57
3	13	62.0	67.8	72.1	27.9	53.8	3
4	1	100.0	100.0	100.0	0.0	0.0	3
5	4	100.0	99.9	99.9	0.1	0.0	6
6	2	100.0	100.0	100.0	0.0	0.0	3
7*	3	100.0	100.0	100.0	0.0	0.0	3
8	7	99.4	96.5	95.8	4.3	0.0	60
9	17	66.3	65.2	69.0	31.0	53.8	3
10*	14	90.2	83.3	85.8	14.3	28.8	12
11	29	84.9	86.8	96.6	3.4	62.5	3
12	9	99.9	95.7	100.0	0.0	25.6	3
13*	7	98.2	93.2	91.9	8.1	0.0	153
14	3	100.0	100.0	100.0	0.0	0.0	3
15	27	75.6	67.2	65.5	34.5	24.4	3
16	5	86.9	87.7	96.9	3.1	58.1	3
17	32	99.5	94.1	92.9	7.1	0.0	72
18	34	94.4	91.7	90.0	10.0	0.0	231
19	2	29.8	75.4	88.5	11.5	90.0	3
20	14	94.5	86.8	85.0	15.0	4.4	45
21*	1	99.7	100.0	100.0	0.0	0.6	3

^*Imply that the end-model is selected for online implementation.