A Privacy Risk Assessment Model for Medical Big Data Based on Adaptive Neuro-Fuzzy Theory

Table 6

Impact of different proportions of illegal users on the model performance.

Proportion of illegal users (%)	N (our model)	Accuracy (our model, Hui et al.’s model, and Wang and Hong’s model)			Recall (Hui et al.’s, Wang and Hong’s, and our model)			F1 value (Hui et al.’s and Wang and Hong’s model)
2.5	15	0.82	0.70	0.69	0.17	0.12	0.12	0.28	0.20	0.20
	30	0.81	0.65	0.64	0.22	0.22	0.21	0.35	0.33	0.32
	45	0.76	0.66	0.63	0.34	0.31	0.28	0.47	0.42	0.39
	60	0.74	0.63	0.61	0.47	0.44	0.34	0.57	0.52	0.44
	75	0.71	0.56	0.55	0.52	0.46	0.41	0.60	0.51	0.47
5	15	0.95	0.90	0.87	0.21	0.17	0.16	0.34	0.29	0.27
	30	0.94	0.89	0.85	0.37	0.33	0.27	0.53	0.48	0.41
	45	0.93	0.87	0.82	0.51	0.47	0.33	0.65	0.61	0.47
	60	0.91	0.85	0.81	0.61	0.56	0.41	0.73	0.68	0.54
	75	0.89	0.83	0.74	0.72	0.67	0.48	0.80	0.74	0.58
7.5	15	0.98	0.95	0.91	0.27	0.19	0.17	0.42	0.32	0.29
	30	0.96	0.93	0.89	0.39	0.32	0.29	0.55	0.48	0.44
	45	0.95	0.92	0.87	0.51	0.46	0.38	0.66	0.61	0.53
	60	0.92	0.90	0.85	0.69	0.64	0.49	0.79	0.75	0.62
	75	0.91	0.89	0.84	0.81	0.79	0.67	0.86	0.84	0.75
10	15	1.00	1.00	0.97	0.27	0.21	0.21	0.43	0.35	0.35
	30	1.00	1.00	0.96	0.39	0.33	0.33	0.56	0.50	0.49
	45	1.00	0.97	0.94	0.59	0.55	0.54	0.74	0.70	0.69
	60	1.00	0.96	0.92	0.72	0.67	0.65	0.84	0.79	0.76
	75	0.97	0.95	0.90	0.83	0.81	0.77	0.89	0.87	0.83
12.5	15	1.00	1.00	1.00	0.26	0.21	0.21	0.41	0.35	0.35
	30	1.00	1.00	1.00	0.43	0.41	0.38	0.60	0.58	0.55
	45	1.00	1.00	0.98	0.60	0.56	0.52	0.75	0.72	0.68
	60	1.00	1.00	0.97	0.76	0.69	0.66	0.86	0.82	0.79
	75	1.00	0.98	0.94	0.83	0.81	0.78	0.91	0.89	0.85
15	15	1.00	1.00	1.00	0.27	0.23	0.22	0.43	0.37	0.36
	30	1.00	1.00	1.00	0.49	0.45	0.39	0.66	0.62	0.56
	45	1.00	1.00	1.00	0.64	0.59	0.54	0.78	0.74	0.70
	60	1.00	1.00	0.98	0.78	0.72	0.68	0.88	0.84	0.80
	75	1.00	1.00	0.95	0.89	0.82	0.76	0.94	0.90	0.84