A Deep Learning Anomaly Detection Framework for Satellite Telemetry with Fake Anomalies

Table 2

DDMN and the compared methods in precision, recall, and F1-score.

(a)

	DDMN
	Precision	Recall	F1-score	Precision	Recall	F1-score
2	1.0	0.939	0.969	0.34	0.947	0.501
2.5	1.0	0.936	0.967	0.654	0.946	0.773
3	1.0	0.928	0.963	0.654	0.945	0.773
3.5	1.0	0.919	0.956	0.654	0.923	0.766
4	1.0	0.9	0.951	0.654	0.907	0.76

(b)

				-SCORE
	Precision	Recall	F1-score	Precision	Recall	F1-score
2	0.342	0.953	0.504	0.707	0.891	0.788
2.5	0.653	0.951	0.774	0.706	0.89	0.787
3	0.653	0.946	0.773	0.814	0.837	0.825
3.5	0.653	0.924	0.765	0.814	0.837	0.825
4	0.653	0.908	0.760	0.814	0.837	0.825

(c)

GMM			-means
Precision	Recall	F1-score	Precision	Recall	F1-score
0.668	0.547	0.601	0.41	0.447	0.428