Differentially Private Kernel Support Vector Machines Based on the Exponential and Laplace Hybrid Mechanism

Table 2

The performance of the DPKSVMEL algorithm on dataset Australian.

k	ɛ	Similarity	Accuracy				AUC
			Mean	Std	Max	Min	Mean	Std	Max	Min
2	0.1	0.5	83.43	2.54	85.94	78.26	0.8963	0.0237	0.9188	0.8352
		0.6	83.36	1.84	86.09	81.30	0.8988	0.0096	0.9148	0.8884
		0.7	85.59	1.15	87.68	83.77	0.9077	0.0211	0.9268	0.8505
		0.8	84.41	2.88	87.25	77.68	0.9162	0.0102	0.9277	0.8992
		0.9	86.07	0.63	87.39	85.22	0.9135	0.0100	0.9246	0.8940
2	0.5	0.5	84.17	2.47	87.10	80.58	0.9041	0.0173	0.9266	0.8787
		0.6	85.32	1.35	87.25	83.04	0.9096	0.0164	0.9297	0.8729
		0.7	84.71	3.59	87.83	75.65	0.9163	0.0101	0.9281	0.8986
		0.8	85.67	1.74	87.97	82.61	0.9220	0.0052	0.9297	0.9116
		0.9	85.88	1.02	87.10	83.48	0.9188	0.0127	0.9324	0.8915
2	1	0.5	85.30	1.83	86.81	80.87	0.9171	0.0071	0.9261	0.9038
		0.6	86.59	1.00	88.12	85.07	0.9258	0.0075	0.9370	0.9122
		0.7	86.23	0.78	87.68	84.78	0.9226	0.0097	0.9317	0.8985
		0.8	86.59	0.79	87.54	85.51	0.9295	0.0026	0.9338	0.9246
		0.9	86.52	0.56	87.39	85.80	0.9287	0.0062	0.9372	0.9181
3	0.1	0.5	80.71	5.92	86.67	68.26	0.8776	0.0410	0.9187	0.7975
		0.6	83.55	2.72	86.96	78.26	0.9023	0.0118	0.9174	0.8849
		0.7	82.65	2.91	86.23	78.41	0.9087	0.0155	0.9240	0.8693
		0.8	85.43	1.48	87.10	82.61	0.9130	0.0135	0.9271	0.8944
		0.9	86.03	0.88	87.39	84.49	0.9078	0.0139	0.9223	0.8764
3	0.5	0.5	81.04	5.49	86.52	68.84	0.8987	0.0197	0.9228	0.8606
		0.6	83.26	2.87	86.52	78.84	0.8994	0.0163	0.9175	0.8711
		0.7	82.20	6.73	87.83	67.25	0.9072	0.0165	0.9277	0.8725
		0.8	85.64	2.14	87.54	80.29	0.9153	0.0132	0.9318	0.8848
		0.9	85.87	1.50	87.10	81.88	0.9224	0.0072	0.9324	0.9096
3	1	0.5	84.87	2.28	87.25	79.57	0.9229	0.0058	0.9297	0.9148
		0.6	85.36	1.42	86.96	82.46	0.9158	0.0099	0.9316	0.8955
		0.7	86.14	1.82	87.54	81.59	0.9230	0.0081	0.9361	0.9105
		0.8	86.20	0.61	87.10	85.22	0.9262	0.0084	0.9368	0.9142
		0.9	86.41	0.57	87.25	85.65	0.9283	0.0040	0.9338	0.9225