Feature Extraction Method Based on Sparse Autoencoder for Air Traffic Management System Security Situation Awareness

Table 1

Parameter settings.

Parameter	Value
Feature association threshold	0.9
a	0.05
λ	0.0005
μ	1
Dimensional change of encoding layer 1	8 ⟶ 7
Dimensional change of encoding layer 2	7 ⟶ 4
Dimensional change of decoding layer 1	4 ⟶ 7
Dimensional change of decoding layer 2	7 ⟶ 8
Penalty coefficient of SVM objective function	0.5