Research Article

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 setsOriginal dataThe comparison Algorithm 1The comparison Algorithm 2The comparison Algorithm 3The algorithm proposed in this paper

wine0.9011 ± 0.01700.9321 ± 0.00870.9493 ± 0.00470.9254 ± 0.00680.9551±0.0053
sonar0.8164 ± 0.01910.8437 ± 0.01870.8608±0.01750.8237 ± 0.01630.8498 ± 0.0139
iono0.8493 ± 0.00620.8988 ± 0.00680.9183 ± 0.00490.8841 ± 0.00590.9268±0.0048
wdbc0.9378 ± 0.00460.9468 ± 0.00620.9560 ± 0.00570.9424 ± 0.00390.9698±0.0031
biodeg0.8250 ± 0.00580.8328 ± 0.00460.8273 ± 0.00510.8257 ± 0.00380.8378±0.0040
messidor0.8273 ± 0.00440.8562 ± 0.00420.8476 ± 0.00490.8455 ± 0.00570.8693±0.0050
winequality-red0.7563 ± 0.00820.8015±0.00740.7962 ± 0.00690.7760 ± 0.00560.7927 ± 0.0053
winequality-white0.7598 ± 0.00390.8019 ± 0.00330.8250±0.00450.7828 ± 0.00480.8056 ± 0.0042
magic0.8019 ± 0.00220.8278 ± 0.00170.8352 ± 0.00360.8159 ± 0.00330.8434±0.0029
average0.8305 ± 0.00790.8601 ± 0.00670.8684 ± 0.00640.8468 ± 0.00620.8790±0.0053

The best experimental results are highlighted in bold.