Anomaly Detection for Aviation Safety Based on an Improved KPCA Algorithm

Table 2

Detection performance by the improved KPCA method.

Training data		Test sample number	Best Parameters	DR (%)	FNR (%)	FPR (%)
Normal sample number	Abnormal sample number
600	266	10000	, ,	93.52	6.48	1.16
800	266	10000	, ,	93.77	6.23	0.91
1000	266	10000	, ,	93.57	6.43	1.11
1200	266	10000	, ,	93.41	6.59	1.27
1400	266	10000	, ,	93.47	6.53	1.21
1600	266	10000	, ,	94.34	5.66	0.34