Research Article

Research on DUAL-ADGAN Model for Anomaly Detection Method in Time-Series Data

Table 2

Comparison of the performance of each anomaly detection model.

ModelRealAdExchange-CPCRealTraffic-SPEEDRealTraffic-TravelTimeAvg
AccPreRecallF1AccPreRecallF1AccPreRecallF1F1

K-means [17]0.7370.4070.730.5230.8570.4740.60.5290.8630.640.6660.6530.568
OC-SVM [20]0.8760.5710.8340.6780.8920.5880.6670.6250.8870.6110.7850.6870.663
LOF [16]0.8840.6250.7690.6890.9110.6920.60.6420.90.6530.80.7190.683
IF [19]0.9080.7850.7330.7580.9370.6660.80.7270.9390.7440.8420.7900.758
LSTM-AE [22]0.9060.750.7740.7610.9380.7850.7340.7580.8830.6330.6790.6550.725
NSIBF [24]0.960.920.8510.8850.9270.880.8150.8460.960.8920.8460.8680.866
Mad-GAN [29]0.9150.720.750.7350.9460.8460.7340.7860.9320.7830.7630.7730.765
Fence-GAN [34]0.9140.710.730.720.9290.78570.6880.7340.940.8480.7370.7890.748
Tad-GAN [30]0.9420.80.8890.8420.9380.7890.8340.810.9380.8280.7630.7950.816
DUAL-ADGAN0.9620.8480.9670.9030.9490.8180.90.8570.960.8450.9260.8830.881