Recent Progress of Anomaly Detection

Table 5

AUC of the anomaly detection algorithms with k=50 for the neighbours, where the performance on Arcene, ALLAML, DLBCL, Gisette, Lung_MPM, and Ovarian was not given, for it was still unavailable after three hours’ running.

Dataset	kNN	ODIN	LOF	LoOP	RBDA	OR	SOD	FB	HiCS
ALOI	0.59	0.75	0.74	0.77	/	0.57	0.71	/	/
Ionoshere	0.48	0.50	0.55	0.63	0.89	0.24	0.77	0.86	0.75
KDDCup99	0.67	0.66	0.62	0.65	/	/	0.89	/	/
PenDigits	0.65	0.66	0.76	0.84	0.92	0.47	0.88	0.96	0.86
Sonar	0.56	0.55	0.55	0.58	0.59	0.49	0.56	0.54	0.61
WDBC	0.61	0.51	0.61	0.50	0.90	0.96	0.90	0.92	0.98
Waveform	0.48	0.48	0.48	0.51	0.73	0.57	0.63	0.73	0.74
Arrhythmia	0.75	0.71	0.74	0.73	0.74	0.68	0.72	0.75	0.61
Ann-thyroid	0.51	0.52	0.53	0.51	0.66	0.54	0.47	0.65	0.52
HeartDisease	0.53	0.47	0.46	0.57	0.56	0.55	0.60	0.64	0.46
Pima	0.57	0.53	0.52	0.62	0.62	0.54	0.65	0.62	0.61
SpamBase	0.63	0.50	0.52	0.53	0.50	0.46	0.55	0.39	0.54