Skip to main content

View full-text article in PMC

. 2022 May 27;24(6):759. doi: 10.3390/e24060759

Table 2.

Experimental results on the MSL, SMAP and SMD datasets. The top 2 F1 scores are bolded.

Method	MSL
Method	Precision	Recall	F1 Score
DAGMM	0.5412	0.9934	0.7007
MSCRED	0.8912	0.9862	0.9363
USAD	0.7949	0.9912	0.8822
MTAD-GAT	0.7917	0.9824	0.8767
OmniAnomaly	0.8867	0.9117	0.8989
GDN	0.9308	0.9892	0.9591
MAD-GAN	0.8517	0.8991	0.8747
GTAD	0.9668	0.9413	0.9539
Method	SMAP
Method	Precision	Recall	F1 Score
DAGMM	0.5845	0.9058	0.7105
MSCRED	0.8175	0.9216	0.8664
USAD	0.7480	0.9627	0.8419
MTAD-GAT	0.7991	0.9991	0.8880
OmniAnomaly	0.7416	0.9776	0.8434
GDN	0.7480	0.9891	0.8518
MAD-GAN	0.8049	0.8214	0.8131
GTAD	0.9821	0.9426	0.9620
Method	SMD
Method	Precision	Recall	F1 Score
DAGMM	0.9869	0.8174	0.8942
MSCRED	0.8164	0.7261	0.7686
USAD	0.9858	0.8174	0.8937
MTAD-GAT	0.7609	0.9999	0.8643
OmniAnomaly	0.8854	0.8827	0.8531
GDN	0.7754	0.7286	0.7513
MAD-GAN	0.9750	0.8827	0.9265
GTAD	0.9515	0.9690	0.9601