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

Table 4

Comparison of kNN 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	0.9011 ± 0.0170	0.9321 ± 0.0087	0.9493 ± 0.0047	0.9254 ± 0.0068	0.9551 ± 0.0053
sonar	0.8164 ± 0.0191	0.8437 ± 0.0187	0.8608 ± 0.0175	0.8237 ± 0.0163	0.8498 ± 0.0139
iono	0.8493 ± 0.0062	0.8988 ± 0.0068	0.9183 ± 0.0049	0.8841 ± 0.0059	0.9268 ± 0.0048
wdbc	0.9378 ± 0.0046	0.9468 ± 0.0062	0.9560 ± 0.0057	0.9424 ± 0.0039	0.9698 ± 0.0031
biodeg	0.8250 ± 0.0058	0.8328 ± 0.0046	0.8273 ± 0.0051	0.8257 ± 0.0038	0.8378 ± 0.0040
messidor	0.8273 ± 0.0044	0.8562 ± 0.0042	0.8476 ± 0.0049	0.8455 ± 0.0057	0.8693 ± 0.0050
winequality-red	0.7563 ± 0.0082	0.8015 ± 0.0074	0.7962 ± 0.0069	0.7760 ± 0.0056	0.7927 ± 0.0053
winequality-white	0.7598 ± 0.0039	0.8019 ± 0.0033	0.8250 ± 0.0045	0.7828 ± 0.0048	0.8056 ± 0.0042
magic	0.8019 ± 0.0022	0.8278 ± 0.0017	0.8352 ± 0.0036	0.8159 ± 0.0033	0.8434 ± 0.0029
average	0.8305 ± 0.0079	0.8601 ± 0.0067	0.8684 ± 0.0064	0.8468 ± 0.0062	0.8790 ± 0.0053

The best experimental results are highlighted in bold.