Research Article

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

Table 9

The performance of the DPKSVMEL algorithm on dataset Splice.

kɛSimilarityAccuracyAUC
MeanStdMaxMinMeanStdMaxMin

20.10.566.735.1774.0058.300.85320.03380.88860.7796
0.670.324.8177.1063.500.84990.00970.86540.8368
0.776.364.1582.5067.200.87890.01770.90630.8512
0.879.566.4685.0068.300.92120.00960.93580.9064
0.985.573.6590.0077.800.95620.00430.96160.9507

20.50.575.024.1083.0069.900.90310.01050.91880.8881
0.677.144.9684.1070.400.89900.01470.91720.8738
0.782.962.5785.7077.400.93010.00800.93780.9127
0.884.236.3189.4068.000.95290.00410.96080.9464
0.989.171.4390.4086.300.97030.00320.97560.9643

210.585.981.3389.3084.400.94250.00730.95540.9312
0.685.353.3088.8080.100.95530.00560.96620.9482
0.788.092.3190.6084.000.96320.00460.96980.9557
0.890.501.2892.0087.800.97290.00360.97760.9644
0.992.170.6192.8091.100.97880.00190.98130.9758

30.10.561.685.2367.9052.800.81500.02390.84670.7750
0.668.395.2075.6057.900.83390.02970.88200.7950
0.774.175.7282.4065.000.87820.01710.90900.8592
0.878.963.3684.7074.000.91510.01070.93430.9009
0.986.623.0590.3082.000.95700.00470.96210.9470

30.50.574.065.7080.0060.800.89240.01880.91270.8488
0.676.096.4581.3060.400.89540.02380.91660.8457
0.779.576.4984.8065.200.92500.00970.93900.9124
0.886.571.8688.8082.800.95050.00500.95810.9443
0.988.844.3491.4076.700.97060.00240.97370.9652

310.583.482.4686.5078.100.93950.00600.94870.9304
0.683.555.8188.7073.700.95420.00470.96000.9449
0.789.001.7090.8085.100.96260.00320.96700.9577
0.891.161.0793.0089.800.97410.00230.97930.9717
0.992.540.8593.5090.900.97930.00130.98150.9769