Uncertainty Measurement and Attribute Reduction Algorithm Based on Kernel Similarity Rough Set Model

Table 3

Comparison of SVM classification accuracy of attribute reduction results

Data sets	Original data	The comparison Algorithm 1	The comparison Algorithm 2	The comparison Algorithm 3	The algorithm proposed in this paper
wine	0.9250 ± 0.0027	0.9493 ± 0.0064	0.9526 ± 0.0057	0.9372 ± 0.0027	0.9657 ± 0.0019
sonar	0.8066 ± 0.0090	0.8550 ± 0.0172	0.8425 ± 0.0145	0.8237 ± 0.0126	0.8564 ± 0.0132
iono	0.8109 ± 0.0043	0.8432 ± 0.0043	0.8521 ± 0.0048	0.8329 ± 0.0062	0.8570 ± 0.0057
wdbc	0.9251 ± 0.0022	0.9572 ± 0.0029	0.9735 ± 0.0045	0.9356 ± 0.0018	0.9697 ± 0.0023
biodeg	0.8250 ± 0.0027	0.8383 ± 0.0049	0.8551 ± 0.0033	0.8290 ± 0.0027	0.8478 ± 0.0036
messidor	0.8458 ± 0.0061	0.8557 ± 0.0067	0.8608 ± 0.0058	0.8593 ± 0.0069	0.8713 ± 0.0062
winequality-red	0.7671 ± 0.0074	0.8129 ± 0.0081	0.7864 ± 0.0062	0.7781 ± 0.0058	0.7932 ± 0.0051
winequality-white	0.7484 ± 0.0017	0.8364 ± 0.0077	0.8242 ± 0.0042	0.7992 ± 0.0065	0.8425 ± 0.0047
magic	0.8266 ± 0.0011	0.8596 ± 0.0023	0.8691 ± 0.0031	0.8431 ± 0.0026	0.8793 ± 0.0018
average	0.8311 ± 0.0041	0.8675 ± 0.0067	0.8684 ± 0.0057	0.8486 ± 0.0053	0.8758 ± 0.0049

The best experimental results are highlighted in bold.