Research Article

A Hybrid Semi-Supervised Anomaly Detection Model for High-Dimensional Data

Table 2

Average AUC and corresponding standard deviation of the different methods. The first class listed in the bracket indicates the normal class. Results were calculated from 50 iterations.

SVDDOCSVMaK-LPEOCSVM-FAOur model

OAR 
(Stand versus others)
AUC 
0.94 
±0.01
0.95 
±0.01
0.96 
±0.00
0.98 
±0.02
0.99
±0.00

OAR 
(Sit versus others)
AUC 
0.88 
±0.01
0.87 
±0.02
0.90 
±0.04
0.90 
±0.01
0.95
±0.01

OAR 
(Lie versus others)
AUC 
0.95 
±0.01
0.95 
±0.01
0.97 
±0.00
0.98 
±0.01
0.99
±0.00

GAS 
(Ethanol versus others)
AUC 
0.92 
±0.02
0.92 
±0.02
0.94 
±0.01
0.93 
±0.02
0.97
±0.01

GAS 
(Ethylene versus others)
AUC 
0.91 
±0.04
0.91 
±0.04
0.92 
±0.03
0.94 
±0.02
0.98
±0.01

GAS 
(Ammonia versus others)
AUC 
0.92 
±0.02
0.93 
±0.02
0.95 
±0.01
0.96 
±0.01
0.99
±0.00

GAS 
(Acetone versus others)
AUC 
0.81 
±0.03
0.77 
±0.04
0.80 
±0.03
0.80 
±0.03
0.92
±0.01

MPID 
(background versus signal)
AUC 
0.70 
±0.04
0.77 
±0.03
0.75 
±0.03
0.79 
±0.03
0.82
±0.01

MPID 
(signal versus background)
AUC 
0.66 
±0.09
0.72 
±0.11
0.73 
±0.03
0.70 
±0.03
0.76
±0.01

KDD 2008 
(benign versus malignant)
AUC 
0.50 
±0.03
0.51 
±0.03
0.34 
±0.01
0.50 
±0.01
0.52
±0.01

Rank4.33.903.102.651.05